Origins of Magnetite Nanocrystals in Martian Meteorite ALH84001

Home , Magnetite, Martian meteorite, Noachian

Author's personal copy

Available online at www.sciencedirect.com

Geochimica et Cosmochimica Acta 73 (2009) 6631–6677 www.elsevier.com/locate/gca

Origins of magnetite nanocrystals in Martian meteorite ALH84001

K.L. Thomas-Keprta a,*, S.J. Clemett a,*, D.S. McKay b, E.K. Gibson b, S.J. Wentworth a

a ESCG at NASA/Johnson Space Center, Houston, TX 77058, USA b KR, ARES, NASA/Johnson Space Center, Houston, TX 77058, USA

Received 20 July 2008; accepted in revised form 18 May 2009; available online 16 June 2009

Abstract

The Martian meteorite ALH84001 preserves evidence of interaction with aqueous fluids while on Mars in the form of microscopic carbonate disks. These carbonate disks are believed to have precipitated 3.9 Ga ago at beginning of the Noachian epoch on Mars during which both the oldest extant Martian surfaces were formed, and perhaps the earliest global oceans. Intimately associated within and throughout these carbonate disks are nanocrystal magnetites (Fe3O4) with unusual chemical and physical properties, whose origins have become the source of considerable debate. One group of hypotheses argues that these magnetites are the product of partial thermal decomposition of the host carbonate. Alternatively, the origins of magnetite and carbonate may be unrelated; that is, from the perspective of the carbonate the magnetite is allochthonous. For example, the magnetites might have already been present in the aqueous fluids from which the carbonates were believed to have been deposited. We have sought to resolve between these hypotheses through the detailed characterization of the compositional and structural relationships of the carbonate disks and associated magnetites with the orthopyroxene matrix in which they are embedded. Extensive use of focused ion beam milling techniques has been utilized for sample preparation. We then compared our observations with those from experimental thermal decomposition studies of sideritic carbonates under a range of plausible geological heating scenarios. We conclude that the vast majority of the nanocrystal magnetites present in the carbonate disks could not have formed by any of the currently proposed thermal decomposition scenarios. Instead, we find there is considerable evidence in support of an alternative allochthonous origin for the magnetite unrelated to any shock or thermal processing of the carbonates. Ó 2009 Published by Elsevier Ltd.

1. INTRODUCTION The majority of these carbonates appear as circular or elliptical features, which range from 10 to 300 lm along the The Allan Hills 84001 (ALH84001) meteorite is a coarse major axis (i.e., eccentricity >0.5). When viewed optically grained orthopyroxenite that solidified from a melt within an exposed fracture surface they range in color from 4.51 Ga ago (Nyquist et al., 2001). Approximately gold to burnt orange in their centers and are typically sur- 600 Ma later (Borg et al., 1999), secondary carbonates were rounded by a thin black–white–black rim (Fig. 1A). The deposited by low temperature hydrothermal processes in three dimensional morphology of these carbonates can best pre-existing fractures and fissures within the groundmass. be described as an inverted conic frustum in which the base of the conic lies flush with the orthopyroxene (OPX) surface and the apex lies under the exposed surface (Fig. 1B). For * Corresponding authors. Tel.: +1 281 483 5029. reasons of brevity we will subsequently user the simpler, al- E-mail addresses: kathie.thomas-keprta-1@nasa.gov (K.L. beit less accurate, description of ‘disk’ when referring to Thomas-Keprta), [email protected] (S.J. Clemett). these carbonates.

0016-7037/$ - see front matter Ó 2009 Published by Elsevier Ltd. doi:10.1016/j.gca.2009.05.064 Author's personal copy

6632 K.L. Thomas-Keprta et al. / Geochimica et Cosmochimica Acta 73 (2009) 6631–6677

Fig. 1A. Optical images of ALH84001 disk-like carbonates. These disks precipitated in fractures produced during shock impacts while the meteorite was on Mars. The majority of these carbonates appear as circular or elliptical features which range from 10 to 300 lm along the major axis. Visually they have colors that vary from gold to burnt orange in their centers which are typically surrounded by a thin black– white–black rim. Some carbonates occur as single entities while others are spatially associated and occur in groups of 10s of carbonates (image at far right). The black rims typically range from 5–10 lm in thickness and are composed primarily of nanophase magnetite embedded in a matrix of Mg-bearing sideritic carbonate while the white bands, typically 10–15 lm thick, are composed of nearly pure magnesite with minor Ca and nanophase magnetite.

thin rim. Embedded within these carbonates are discrete nanocrystal magnetites (Fe3O4) that exhibit super-para- magnetic to single-domain properties (Muxworthy and Wil- liams, 2008). While the largest population of these magnetites occurs within the rim zone of the carbonate disks, e.g., McKay et al. (1996) and Thomas-Keprta et al. (2000a), a small but significant fraction is also found within the inner and outer core zones. When the carbonate disks were initially characterized, they were interpreted as products of high temperature processes (T > 573 K) (Harvey and McSween, 1996; Mit- tlefehldt, 1994a; Mittlefehldt, 1994b; Scott et al., 1997). Subsequently, isotopic and chemical data has forced a revi- sion in this interpretation and it is now generally accepted they were deposited by low temperature (T < 573 K) hydro- Fig. 1B. Illustration of an inverted conic frustum, that is, the thermal processes (Eiler et al., 2002; Kent et al., 2001; Kirs- polygon which remains after a cone is cut by a plane parallel to the chvink et al., 1997; McSween and Harvey, 1998; Romanek base with the apical part removed and the remaining polygon et al., 1994; Saxton et al., 1998; Treiman, 1997; Valley et al., inverted. While single ALH84001 carbonates have been loosely described as rosettes, globules, concretions, or pancakes, they are 1997; Warren, 1998). While the consensus of opinion on the more accurately described as being shaped like an inverted conical temperature of formation of the carbonates has converged, frustum. that of their subsequent thermal evolution has not because of the differing hypotheses invoked to explain the presence of the nanocrystal magnetites present within the carbonate. Chemically the disk carbonates have a complex mixed To first order, these magnetites were either already present cation composition with an empirical formula (FexMgy- when the carbonates formed and so became embedded dur- CazMn[1xyz])CO3 where x + y + z = 1. The values of ing carbonate deposition, or they were formed a posteriori x, y, and z vary with radial distance in such a way that by partial thermal decomposition of the pre-existing car- the carbonate disks can be envisioned as being composed bonate. In the latter case, the heating of the meteorite is of three concentric annular zones; starting from the center envisioned to have occurred through the adiabatic propaga- there is an inner central and outer core surrounded by a tion of one or more impact shock events. Author's personal copy

Origins of Magnetite in ALH84001 6633

ALH84001 shows evidence of having undergone multi- 1996; Thomas-Keprta et al., 2000a; Thomas-Keprta et al., ple impact events, the last resulting in its ejection from 2001; Thomas-Keprta et al., 2002). Mars, e.g., Treiman (1998). In general, the peak shock pressures experienced by Martian meteorites are thought to 2. METHODOLOGY range from less than 14 gigapascals (GPa) up to 55 GPa (Fritz et al., 2004; Fritz et al., 2003; Langenhorst et al., 2.1. ALH84001 carbonate disks 2000; Sto¨ffler, 2000; Treiman, 2003; van der Bogert et al., 1999). These estimates are based on the conversion of pla- Table 1 lists the ALH84001 rock splits used in the re- gioclase to either a diaplectic ‘maskelynite’ (14–45 GPa) search presented herein. Prior to any extraction, fracture or melt glass (>45 GPa) with a resulting increase in the surfaces containing carbonate disks were first characterized refractive index and/or Raman band broadening, compared by optical and electron microscopy. This involved optical to unshocked plagioclase (Fritz et al., 2004). Specifically, imaging under both bright and dark-field illumination con- for ALH4001, estimated values for peak shock pressure ditions. Given the topographic variations present within a have evolved over time – 50–60 GPa (Langenhorst given fracture surface, a long working distance Nikon et al., 2000); <40 GPa (van der Bogert et al., 1999); 35– Eclipse ME600 microscope was used for this task. Due to 40 GPa (Sto¨ffler, 2000); 35.7 ± 4.5 GPa (Fritz et al., the low numerical apertures associated with long working 2003); >35–40 GPa (Fritz et al., 2004) – with the best cur- distance microscope objectives, the depth-of-field was rent estimate at 32 ± 1 GPa (Fritz et al., 2005). The extent to which impact shocks to ALH84001 lead to thermal excursions that could have partially decomposed Table 1 the ALH84001 carbonate disks remains ambiguous (note, Rock splits of Martian meteorite ALH84001. in this context any impacts occurring in the 600 Ma prior Split Parent to carbonate formation have no bearing). Results from sim- 157 21 ulations of shockwave propagation through composition- 158 40 ally and texturally complex matrices are computationally 160 2 difficult and sensitive to initial starting conditions, but do 161 17 indicate that shock heating would have been highly hetero- 162 41 geneous at all size scales. Using a Hugoniot equation-of- 163 69 state experimentally derived from Stillwater pyroxenite, 164 76 165 103 Twin Sister dunite (Sto¨ffler, 1982), and a gabbroic rock 166 147 (Trunin et al., 2001), the post-shock temperature elevation 174 39 for ALH84001 associated with a 32 GPa shock event lies 175 39 in the range of 373–383 K (Fritz et al., 2005). If we as- 180 69 sume ALH4001 prior to ejection was at the average Mar- 181 69 tian ground-surface temperature of 230 K and that the 182 69 peak shock pressure experienced corresponded to this final 184 14 ejection event (note, paleomagnetic data of Weiss et al. 185 14 (2000) suggest the interior of ALH84001 was not heated 186 14 above 313 K), this would imply an average maximum 188 69 189 26 heating, post-ejection, to a temperature of 610 K. This 190 147 is in broad agreement with recent (U–Th)/He analyses of 191 128 ALH84001 phosphate grains that suggest a peak ejection 192 34 temperature of 673 K or less (Min and Reiners, 2006). 193 134 To address these unresolved questions we have utilized 195 59 focused ion beaming techniques in conjunction with con- 197 40 ventional ultramicrotomy to provide the most detailed pic- 198 0 ture yet of the spatial, structural and chemical relationships 200 0 present within ALH84001 carbonate disks. We have then 208 14 used these new observations to evaluate the merits of both 217 69 283 214 theoretical and experimental attempts to reproduce 284 208 ALH84001-like disks with embedded magnetites through 286 218 partial thermal decomposition of Fe-bearing carbonate. 289 202 In the larger context, resolving or at least constraining the 297 213 origin of the magnetites within ALH84001 carbonate disks 372 239 has importance in that it has been previously suggested that 374 103 the precipitation of the carbonate disks may have been 383 66 facilitated through biogenic activity (McKay et al., 1996) 384 61 and that a fraction of the embedded magnetites exhibit 385 104 chemical and physical features identical to those produced 386 247 388 153 by contemporary magnetotactic bacteria (McKay et al., Author's personal copy

6634 K.L. Thomas-Keprta et al. / Geochimica et Cosmochimica Acta 73 (2009) 6631–6677 generally insufficient for any single image to be in complete Table 2 focus. We therefore acquired image stacks over a range of Split number and size of ALH84001 focused ion beam sections. focus planes for each imaged sample surface, which were Carbonate section Split number FIB extracted section then subsequently combined using a complex-wavelet size (lm) algorithm (Forster et al., 2004) to generate a single in-focus Texas, Section 1 ,286 15 8 image with an extended depth-of-field. After optical imag- Texas, Section 2 ,286 17 5 ing, selected fracture surfaces were transferred to 1/2-inch Posterboy, Section 1 ,386 17 10 Al pin mounts and attached using conductive C tape before Posterboy, Section 2 ,386 16 11 being lightly coated with either C or Pt (<1 nm) to enable Posterboy, Section 3 ,386 15 9 imaging using a JEOL 6340F field-emission scanning elec- Posterboy, Section 4 ,386 13 8 tron microscope (SEM) at 15 kV. In one case, a sample chip Posterboy, Section 5 ,386 15 10 containing a large, elongated (250 115 lm) carbonate disk (a.k.a ‘Ear’ carbonate) was also analyzed using a face tension between and across adjacent faces. In practical Cameca SX100 electron probe microanalyzer at 15 kV terms for 150 nm thick carbonate thin sections we found and 10 nA to determine both major and minor element the length-to-thickness (l/w) ratio could not practically ex- compositions. Both quantitative linescans and qualitative ceed 200, with the largest FIB section extracted here being 2-D element maps were collected at a lateral spatial resolu- 20 lm in length (l/w 130). tion of 1 lm for C (dolomite standard), Mg (kaersutite In all, seven FIB thin sections were acquired using the standard), Si (kaersutite standard), S (troilite standard), standard ‘H’-bar extraction procedure from two carbonates Ca (calcite standard), and Fe (kaersutite standard). on unrelated fracture surfaces as documented in Table 2. After the initial in situ characterization, parts of selected Sections were prepared using either a FEI Dual-Beam Stra- carbonates were extracted for analysis at higher resolution ta 237 (‘Texas’ Sections 1 and 2; ‘Posterboy’ Sections 1–4) using transmission electron microscopy (TEM) and energy or Strata 400 (‘Posterboy’ Section 5) workstations. Both dispersive X-ray spectroscopy (EDX). Several extraction instruments were equipped with a field emission electron procedures were utilized; at it simplest, this involved the di- source, a Ga+ ion source, a 30 kV scanning TEM detector, rect mechanical separation of either a complete or partial an EDX spectrometer, and an Omniprobe in vacuo micro- carbonate disk from its OPX fracture surface using a pair manipulator. Sample mounting used the same 1/2-inch Al of surgical austenitic stainless steel tweezers. Removed car- pin mounts used in the preliminary SEM imaging. The first bonate fragments were then crushed between a pair of clean two FIB thin sections were acquired from an interior car- glass microscope slides to reduce the average particle size to bonate fragment designated ‘Texas’ plucked from a freshly less than 50 lm. From this crushed material, individual exposed OPX fracture surface of rock split ALH84001,286. fragments of carbonate, typically less than 20 lm in size, Both sections were transferred directly to a continuous C were dry-picked and embedded within a small drop of un- film Cu TEM grid. The remaining five FIB thin sections cured Embed 812 epoxy resin atop an epoxy-potted butt. were extracted from both core and rim of a carbonate disk After curing the carbonate, embedded epoxy samples were designated ‘Posterboy’ located on a freshly exposed fracture sectioned into 70–100 nm thick electron transparent sec- surface of rock split ALH84001,386. Note that for these tions using a Reichert MT 7000 ultramicrotome equipped sections the carbonate was not removed from its fracture with a 45° angle diamond knife. Individual sections were surface, rather the entire chip containing the fracture sur- ‘floated off’ the diamond knife after cutting using a triple face was mounted in the FIB instrument. This was done distilled water bath and transferred to either Cu or Au in order to permit the study of both carbonate and the 200 mesh thin-bar TEM grids. interface with the surrounding and underlying OPX. Each While ultramicrotomy is the de facto preparation tech- removed thin section was welded on one or both sides nique for TEM thin sections, it proved to have two signif- in situ to a Cu cresent TEM mount. icant disadvantages in the case of ALH84001 carbonate Both ultramicrotome and FIB thin sections were ana- disks. First, all essential spatial information is lost between lyzed using either a JEOL 2000 FX 200 kV STEM and/or the final thin section and the source carbonate disk from a JEOL 2500SE 200 kV field emission STEM. Both instru- which it was extracted. Second, the risk exists of potential ments were equipped with light element (Z > 5) Si-drift aqueous alteration and/or contamination when a cut thin EDX detectors (Noran System 6), and in the case of the section is floated-off off the diamond ultramicrotome knife 2500SE, a modified objective-lens pole piece allowed a col- using a water bath. To circumvent these issues, we used a limator-equipped 0.3 steridian large area (50-mm2) EDX second approach for the preparation of carbonate sections, detector for increased sensitivity. The JEOL 2500SE instru- that being a focused ion beam (FIB) instrument (Giannuzzi ment was also equipped with a Gatan Tridiem Imaging Fil- and Stevie, 1999). In the FIB technique, a tightly focused ter for energy-filtered imaging and electron energy-loss ion beam is used to selectively cut a TEM section directly spectroscopy, and a high-angle annular darkfield detector from the surface of a sample; this permits site specific for z-contrast imaging. extraction with complete preservation of spatial location. Furthermore, FIB section preparation introduces very little 2.2. Roxbury siderite mechanical stress, so that the size of the thin section that can be obtained is limited only by the tendency of large area To investigate the chemical compositions of magnetite surfaces to eventually warp or curl due to differences in sur- formed from the decomposition of Fe-rich carbonates, a Author's personal copy

Origins of Magnetite in ALH84001 6635

186 g sample of siderite (FeCO3) from the Roxbury iron mine in Litchfield, Connecticut was obtained from the Excalibur Mineral Company. Roxbury siderite has previously been reported as having a bulk composition of (Fe0.84Mg0.10Mn0.04 Ca0.02)CO3 (Lane and Christensen, 1997) and provides (compositionally) a reasonable terrestrial analog to the Fe-rich component of ALH84001 carbonates disks, e.g., a composition of (Fe0.750Mg0.240Mn0.003Ca0.004)CO3 was calculated by Treiman (2003) for the inner rim of a carbonate disk. Three separate sample fractions of Roxbury siderite were prepared by mechanical abrasion from the bulk sample. One fraction served as a control while the remaining two were thermally decomposed. In one case, decomposition occurred under a very slow, controlled heating rate, while in the other, the heating event was essentially instantaneous. We will, for brevity, subsequently refer to these heating scenarios as either being ‘slow’or‘fast’, Fig. 1C. Backscatter electron image (BSE) of a partial, false- respectively. colored ALH84001 carbonate disk embedded within the OPX In the ‘slow’ heating experiment the large sealed quartz matrix of a freshly fractured chip. The uppermost surface of the tube (LSQT) method was used to produce a closed reaction carbonate lies flush with the surrounding OPX (gray). The body of under conditions approximating thermal equilibrium (Lau- the carbonate disk is inset within a form-fitting depression 5to er et al., 2005). In this method, the siderite sample is placed 8 lm deep within the OPX. All ALH84001 carbonate disks we have in small open quartz crucible, which is in itself within a lar- investigated have this common morphological context. ger evacuated quartz tube, along with a second crucible containing lime (CaO). The lime serves as a sink for CO2 liberated during carbonate decomposition and so acts to between the adjacent fracture surfaces, but rather, is offset maintain a constant oxygen fugacity ðf Þ. The whole O2 such that the carbonate disk is either mostly or wholly inset assembly is heated in a vertical tube furnace, with the tem- within a form-fitting pit in one or other surface. Intrigu- perature being controlled using a J-type thermocouple at- ingly, in fractures in which multiple carbonates are present, tached to the LSQT. In our experiment, the siderite 1 the pits in which each of the carbonates are inset all lie on sample was heated at a rate of 1 K min to a maximum the same OPX surface. A priori it might be expected that the temperature of 848 K, after being held at this temperature carbonate pits would be evenly distributed between the for 24 h, it was to cooled back to room temperature at a adjacent fracture surfaces, however, this is not so. rate of 102 K s1. In the ‘fast’ heating experiment, the siderite sample was 3.1.2. External shape of ALH84001 carbonate disks heated under vacuum with a pulsed 10.6 lmCO2 laser The carbonate disks within each OPX pit show a rela- (PRF-150 Laser Science, Inc.) focused onto the sample tively constant thickness that does not change, proceeding using a Cassegrainian microscope objective. The temporal from the core outward, until the onset of the inner magne- profile of the CO2 laser pulse had an 300 ns initial spike tite band where upon it abruptly decreases, in an approxi- accounting for 30% of the total pulse energy which was mately linear fashion, all the way to the carbonate edge. then followed by a slow oscillatory decay extending up to Hence, as noted previously, the most accurate terminology 1–2 ls. Results using a thin-film Pt thermocouple suggest 1 8 9 1 describing this shape is that of an inverted conic frustum , that the heating rates were on the order of 10 –10 K s see Fig. 1B, as opposed to the more widely used, but inac- with a peak temperature of 573–673 K being reached at curate description, as disks, rosettes, or pancakes. In gen- the laser focus (Zenobi et al., 1995). eral, while the diameter of carbonate disks can vary by a factor of 10 or more, even within a single fracture surface, 3. RESULTS the thickness is relatively invariant being typically 10– 15 lm. Consequently this implies that the radial thickness 3.1. ALH84001 carbonate disks of the carbonate rims must also be similarly invariant, as is observed with a typical width of 10 lm. 3.1.1. Spatial location and distribution of carbonate disks Carbonates account for 1% by volume of ALH84001 and are distributed throughout the multiple internal fracture surfaces that pervade the meteorite (Mittlefehldt, 1994a; Mittlefehldt, 1994b). While single isolated carbon- 1 In the rim region the inner and outer magnetite bands do not lie ates do occur, most often they are found in clusters ranging perpendicular to the fracture surface but rather are canted outward from a few to tens of carbonates, some of which appear with acute angles / and h, where / > h and h (90 u), with u partially fused (Fig. 1A). These carbonates are not simple being the angle between the generatrix and the axis of the conic vein-filling deposits since their center-of-mass lies not frustum. Author's personal copy

6636 K.L. Thomas-Keprta et al. / Geochimica et Cosmochimica Acta 73 (2009) 6631–6677

Fig. 1D and 1E. SEM images typical of the striated/ridged texture that characterize the OPX fracture surface adjacent to carbonate disks. Such textures could result from ﬂuid dissolution of the OPX surface.

3 4 3.1.3. Morphology, textures, and relationship of carbonate to (Si2Al2O5(OH)4), smectite and serpentine clay minerals, OPX and oxyhydroxides such as goethite (a-FeOOH) (Velbel, Carbonates are inset within form fitting pits and lie flush 2007). While there have been sporadic observations of with the OPX fracture surface as illustrated in Fig. 1C. The smectites in ALH84001 (Thomas-Keprta et al., 2000b), gen- OPX fracture surfaces in which carbonates are found all erally the most remarkable observation is how scarce they demonstrate an aligned striated texture, Fig. 1D and 1E, appear to be. Given the age and uncertainty in the history which can be interpreted as the result of differential weath- of ALH84001, it is possible that such minerals might have ering of OPX lamellae that is characteristic of partial aque- once been present but were not preserved (Velbel, 2007), ous chemical dissolution (Velbel, 2007). This texture nevertheless the lack of such weathering minerals associated contrasts sharply with that of the OPX underlying the car- with ALH84001 fracture surfaces and their associated carbonate which forms the base of the pits, which has a loosely bonate disks has yet to be satisfactorily explained and re- oriented botryoidal or micro-dentricular texture (Fig. 1F mains an enigma. and 1G). These Martian ‘microdenticles’ share some physical similarities to terrestrial denticles formed by the low 3.1.4. Surface textures of carbonate disks temperature aqueous alteration/weathering of chain-sili- SEM images of the upper surfaces of carbonate disks ex- cates2 (e.g., pyroxene, amphibole) (Velbel, 2007; Velbel posed to fracture space are shown in Fig. 1H and 1I. At low et al., 2007). Although the majority of terrestrial denticles magnification, carbonates display a pitted texture inter- are larger in size than their putative Martian counterparts spersed with lamellar striations, while at high magnification (up to 10s of microns in length) the lower end of the size the surface texture is composed of discrete particles ranging distribution does overlap that of the Martian microdenti- from 20 to 250 nm with ovoid to irregular shapes. This cles (Velbel, 2007; Velbel et al., 2007). Hence, this suggests texture is unlikely to represent the primordial texture of that the OPX textures, including pits and etched surfaces, the carbonate at the time of its deposition, but rather rep- formed either prior to, or contemporaneously, with the resents the products of subsequent superficial erosion and deposition of carbonate through a process of chemical dis- carbonate dissolution during post-depositional fluid solution of the silicate. interactions. In the terrestrial context, aqueous weathering of chain- silicates results in the concurrent formation of other 3.1.5. Composition of ALH84001 carbonate disks low temperature secondary minerals such as kaolinite The largest carbonate within a cluster of carbonates on a freshly exposed fracture surface of ALH84001,286 was designated as the ‘Ear’ carbonate (Figs. 2A and 2B). This ‘Ear’ carbonate provides a good example of the typical carbonate

2 3 Pyroxenes have the general formula XY(Si,Al)2O6 (where X The smectite group of minerals includes dioctahedral phyllosil- 2+ 3+ generally represents Na, Mg, Ca, and Fe and Y generally icates including Fe -rich nontronite (Na0.3Fe2(Si,Al)4O10(OH)2 3+ 3+ represents Mg, Al, Sc, Ti, V, Cr, Mn, and Fe ). Amphiboles have nH2O), Fe -poor montmorillonite ((Na,Ca)0.33(Al,Mg)2- chemical compositions and general characteristics that are similar Si4O10(OH)2 nH2O) and trioctahedral clays including saponite 2+ to the pyroxenes although extensive substitution of Si by Al can with the general chemical formula Ca0.25(Mg,Fe )3- occur. The chief diﬀerence between amphiboles and pyroxenes is (Si,Al)4O10(OH)2 4H2O. 4 that the basic structure of an amphibole is a double chain of SiO4 Serpentine represents a group of hydrous Mg, Fe-phyllosilicate tetrahedra compared to a single chain structure for pyroxene. ((Mg,Fe)3Si2O5(OH)4) minerals. Author's personal copy

Origins of Magnetite in ALH84001 6637

Fig. 1F and 1G. SEM views of the typical OPX surface texture that underlies carbonate disks (residual overlying carbonate is shown false colored in orange in 1F), composed of loosely oriented, botryoidal or micro-dentricular features with apexes that range from sharply pointed to rounded. In terrestrial weathered silicates, the nature of such denticulated features is strongly influenced by the composition of the dissolution fluid and its degree of undersaturation relative to the etched surface, see Velbel (2007) and Velbel et al. (2007) and references therein. While the Martian microdenticles are smaller in scale than those observed in terrestrial silicates, there is sufficient similarity in the textures of the Martian and terrestrial denticles to permit the interpretation that aqueous dissolution could account for their presence. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this paper.)

Fig. 1H and 1I. SEM view (H) showing the characteristic deeply ridged and pitted (e.g., honeycomb) textures of the carbonate disk contiguous with the OPX fracture surface. At higher magnification (I) this surface texture is comprised of a matrix of spheroidal, ovoid, and irregular-shaped grains ranging from 10 to 300 nm along the longest axis. Note that this surface texture is independent of the underlying carbonate composition, that is, the magnetite-rich rims display a texture nearly identical to that of the core carbonate. disk. While there are compositional variations between car- obtained by electron microprobe analyses of the ‘Ear’ carbonate disks, both on the same and different fracture sur- bonate, while Table 3 lists the bulk composition. faces, many of the broader characteristics demonstrated While at the one micrometer resolution limit of the elec- by the ‘Ear’ carbonate are representative of all carbonates, tron microprobe the variations of Ca, Mg, Fe, and Mn cat- albeit to differing degrees. Hence we have used the ‘Ear’ ions appear complex and sometimes chaotic, at a broader carbonate as the archetype to describe the spatial and com- scale clear radial symmetric zoning patterns are apparent positional relationships exhibited by carbonate disks5. that can be described as three approximately equivolume Fig. 2A shows quantitative 2-D cation distributions concentric zones6: an inner core (50 lm radius along the major axis; 30 vol.%), surrounded by an outer core

5 For example, the Ca-rich cores characteristic of the larger 6 This delineation into three zones is purely a construct to carbonates are generally absent in those carbonates that populated simplify the description of the carbonates and does not necessarily the lower end of the carbonate size distribution. imply any underlying signiﬁcance. Author's personal copy

6638 K.L. Thomas-Keprta et al. / Geochimica et Cosmochimica Acta 73 (2009) 6631–6677

Fig. 2A. BSE view of a single ALH84001 carbonate disk, designated as ‘Ear’ carbonate and its associated quantitative element distribution maps for Ca, Fe, Mn, Mg, and S, acquired by wavelength dispersive electron microprobe analyses using a 1 lm probe beam. The region we designate as the inner core is characterized a signiﬁcant enrichment in Ca and Mn, with both showing oscillatory, radial zoning patterns. The region that separates the inner core from the rim we designate as the outer core. The transition between the inner and outer core is characterized by a sharp (<1 lm) compositional boundary in Ca and Mn distribution. The major cation distribution (Ca, Mg, Fe, and Mn) in this outer core is radially invariant, that is it changes little with increasing radius. Enrichment of S in the inner and outer magnetite-rich rims is attributed to the presence of nanophase Fe-sulﬁdes.

(20 lm thick radial band; 30 vol.%), which itself is sur- The simplest of the three zones is the rim which is com- rounded by a rim (20 lm thick radial band; 40 vol.%) as posed of a relatively thick band of almost pure magnesite illustrated in Fig. 2B. (MgCO3) (>90 mol.% Mg) which is bordered on its inner Author's personal copy

Origins of Magnetite in ALH84001 6639

and outer perimeter by thin Fe-rich bands composed primarily of magnetite embedded in a Mg:Fe (60:40) carbonate matrix. Within the rim lies the outer and inner cores of which the latter is the compositionally most complex. At its very center, the inner core has a Ca-rich (Ca:Mg:Fe:Mn 45:30:15:10) nucleus (75 20 lm major/minor axis) accounting for 5% of the total volume of the carbonate. This is encircled by a thin (10 lm) Ca-rich halo (Ca:Mg:Fe:Mn 30:30:30:10) separated from the nucleus by a thin band (10 lm) of Mg–Fe-rich carbonate (Mg:Fe:Ca:Mn 40:40:16:4). Beyond the Ca-rich halo, the Ca concentration decreases proportionally to increasing Mg, while Fe concentration remains relatively invariant. Although Mn is always the minor cation, its relative abundance closely correlates with that of the Ca cation distribution, and is loosely inversely correlated to the Fe cation distribution. The outer core which separates the inner core from the rim is characterized by Ca, Mg, Fe, and Mn cation distributions that do not change significantly with changing radius, i.e., they are radially invariant. Another way to visualize the compositional differences between the inner and outer cores and rim is to use the Ca, Mg, Fe + Mn ternary diagram in which the carbonate compositions associated with these three zones can then be seen to form three separate, but partially overlapping, density ellipses (Fig. 2C). For comparison to Fig. 2C, the Ca, Mg, Fe ternary diagram outlining the stability fields for calcite (CaCO3), dolomite (Mg,Ca(CO3)2) – ankerite (Ca,Fe(CO3)2), and the complete magnesite-siderite solid solution series is shown in Fig. 2D. It is apparent that the compositions of many ALH84001 core and rim carbonates lie outside of these thermodynamic stability fields. Com- bined with the complex radial oscillatory zoning and sharp compositional boundaries observed, it is clear that ALH84001 carbonate disks represent assemblages in a high degree of disequilibrium, consistent with the chemical and isotopic findings from previous studies, e.g., Treiman (1997) and Valley et al. (1997). If the carbonate disks evolved Fig. 2B. Upper image: colorized image of the ‘Ear’ carbonate, by successive radial deposition of carbonate from a central 250 115 lm in size, shown in 2A indicating approximate nucleus it would therefore imply that the pore- or fracture- contours of the inner and outer cores, and rim. The solid red line filling fluid was out of thermodynamic equilibrium with the bisecting the carbonate represents the axis along which 151 electron surrounding rock matrix, and that variations in diffusion microprobe spot analyses were acquired. Lower image: cation mole parameters led to the development of the complex deposition fractions for ‘Ear’ carbonate regions analyzed along the solid red patterns. This is commonly observed in environments where line. Transitions between the inner and outer core and rim appear the rate of fluid-rock mass transfer is limited by diffusion. as abrupt changes (<1 lm) in the carbonate cation composition as The observed complex mineral zonation patterns thus record indicated. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this the changes in the fluid composition caused by variable paper.) kinetic dispersion and fluid flow rates at the site of crystal growth. Oscillatory patterns formed where the rate of

Table 3 Electron microprobe analyses of ALH84001 ‘Ear’ carbonate. ALH84001 Min. Max. Ave. 1 Weighted Ave. ALH84001 Carbonate Ave. Roxbury Siderite Ave. Carbonate Disk (mol.%) (mol.%) (mol.%) 241 analyses (mol.%) (Mittlefehldt, (mol.%) (Lane and ‘Ear’ carbonate (mol.%) 1994b) Christensen, 1997) Fe 0.011 0.731 0.336 ± 0.102 0.375 0.294 0.84 Mg 0.028 0.951 0.462 ± 0.147 0.493 0.580 0.10 Ca 0.015 0.526 0.172 ± 0.115 0.114 0.115 0.02 Mn 0.001 0.098 0.030 ± 0.024 0.018 0.011 0.04 Author's personal copy

6640 K.L. Thomas-Keprta et al. / Geochimica et Cosmochimica Acta 73 (2009) 6631–6677 diffusion was sluggish enough to permit the establishment of respectively. TEM overview images of ‘Texas’ FIB Sections sharp stationary chemical potential gradients. 1 and 2, and ‘Posterboy’ FIB Sections 1–5 are shown in Figs. 3C–3I, respectively. Rather than discuss each section 3.2. ALH84001 FIB sections individually, for succinctness Fig. 4 summarizes the underlying characteristics of the rim and core carbonate struc- SEM images of ‘Texas’ and ‘Posterboy’ carbonates prior tures observed in these sections. to and after FIB extraction are shown in Figs. 3A and 3B, 3.2.1. Sections of rim carbonate: ‘Posterboy’ Sections 1–5 Compositionally the inner and outer magnetite rims are composed of an approximately 50:50 vol.% mixture of fine grained carbonate and single crystal magnetites (Figs. 2A and 2B; Fig. 5A), with Fe-sulfides also present as a minor phase (Fig. 5A). The carbonate has a Mg:Fe cation ratio of approximately 40:60 with minor and trace amounts of Mn and Ca, respectively. Pervading this carbonate is a poorly constrained, minor amorphous silica phase (Fig. 5A). Although this silica phase is also present throughout the entire carbonate, it is enriched 3-fold within these rims. In some cases, we have also observed it to constitute a major phase and note that Si-rich veins in ALH84001 carbonates were also described by McKay et al. (1998). The magnetites are predominately stoichometrically pure Fe3O4 (Fig. 5B), although some do contain Cr and/or Al (Fig. 5C) as noted previously by Thomas-Keprta 3 Fig. 2C. Ternary cation plots for ‘Ear’ carbonate based on 449 electron microprobe spot analyses. In each plot the three cation end-members are Ca, Mg, and (Fe + Mn) and each data point is colored according to the mole fraction of Mn. In the top plot all 449 data points are show together while in the lower plots, data points corresponding to the inner and outer core and rims are shown plotted separately. The carbonate compositions associated with these three zones can be seen to form three separate, but partially overlapping, density ellipses. In the rim zone, the major axes of the density ellipse lies perpendicular to the Ca-end member

calcite (CaCO3), encompassing a range of Ca-poor carbonate compositions with varying Fe:Mg ratios. Transitioning to the outer core, the density ellipse that encompasses these carbonates has a composition that, while overlapping the rim density ellipse, nevertheless has a slight positive gradient with respect to the horizontal axis deﬁned by the Mg- and Fe(Mn)-end member

carbonates, magnesite and siderite (rhodochrosite; MnCO3). Hence, the more Fe-rich carbonates tend to be proportionally more Ca-rich, although the total fraction of Ca still remains low. The carbonate composition of the inner core shows a density ellipse that is almost perpendicular to that of both the outer core and rim, deﬁning a range of Ca-rich to Ca-poor carbonates that share similar Mg:Fe ratios. The lower, Ca-poor, end of the density ellipse overlaps with that of the outer core while the upper Ca-rich end

extends up to carbonates with an ankeritic (Ca,Fe)(CO3)2)/ dolomitic (Mg,Ca(CO3)2) composition. The different compositions of the inner and outer cores and rim are reflected in the separate data point clusters associated with each region. In the Ca-poor carbonate of the rim, the (Fe + Mn):Mg ratios are seen to vary nearly over the complete range between end-members. In the outer core, a smaller spread in (Fe + Mn):Mg ratios are observed and the carbonate tends to be proportionally more Ca-rich, although the total fraction of Ca still remains low. Finally in the Ca-rich inner core, the (Fe + Mn):Mg ratios are almost invariant relative to the mole fraction of Ca, which ranges up to almost an ankeritic- dolomitic composition. This results in an elliptical distribution of points with the major axis perpendicular to that defined by the (Fe + Mn) – Mg end members. Author's personal copy

Origins of Magnetite in ALH84001 6641

Fig. 2D. Calculated thermodynamic stability phase diagram for Fe–Mg–Ca carbonates at 750 K and 1 bar pressure, using the approach of McSwiggen (1993a) and McSwiggen (1993b). Not all carbonate compositions are thermodynamically stable resulting in three distinct stability fields, highlighted in blue, corresponding to calcite, dolomite–ankerite, and the magnesite–siderite solid solution series. While the compositional range of ALH84001 carbonate disks extends into both the magnesite–siderite (2C, rim carbonate) and ankerite–dolomite stability fields (2C; inner core carbonate), most carbonate compositions lie between these stability fields indicating that for any given carbonate disk, taken as a whole, is in a state of thermodynamic disequilibrium. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this paper.)

Fig. 3B. Optical (upper) and BSE (lower) views of ALH84001 carbonate disk designated as ‘PosterBoy.’ In the upper optical image the carbonate is shown prior to FIB extraction (red boxes) while the lower BSE image was acquired after ﬁve separate FIB extractions, four of the rim and one of the core. (For interpretation of the references to color in this ﬁgure legend, the reader is referred to the web version of this paper.)

Morphologically the inner and outer cores and magnesite band are cross-cut by numerous veins (Figs. 3E, 3H and 3I, Figs. 5A and 5D) and open fractures (Fig. 3I), respectively, which must have formed after initial carbonate deposition. In the case of the magnesite band (Fig. 3I) most fractures appear open and are oriented along grain boundaries. A small percentage appears to be partially filled Fig. 3A. BSE view of a partial ALH84001 carbonate disk with an amorphous Fe phase(s) (Fig. 3I) that we have not designated ‘Texas’ that was removed from the OPX fracture yet been able to characterize. In the case of the outer core surface in which it was embedded and transferred to electron carbonate, there is extensive penetration by veins that radi- conductive double-sided C tape (black background). The surface ate inward toward the inner core, either horizontally from shown corresponds to the exposed surface which was contiguous the inner magnetite rim, or vertically from the upper car- with the OPX. Locations of the two core carbonate FIB sections bonate surface (Figs. 3E, 3H and 3I, Figs. 5A and 5D). that were extracted are shown by the boxes. These veins are completely filled with a fine grained carbonate compositionally similar to the carbonate into which it intrudes. However, unlike the surrounding carbonate, the et al. (2000a). Sandwiched between the inner and outer vein filling is intimately mixed with one or more amorphous magnetite bands is the almost pure magnesite band contain- silica phases (Figs. 5A and 5D) enriched by a factor of 3 ing minor Ca and Si (Fig. 3G) and sporadic magnetite crys- with respect to the surrounding carbonate, and contains tals (Fig. 3G). embedded stoichometrically pure single crystal magnetites Author's personal copy

6642 K.L. Thomas-Keprta et al. / Geochimica et Cosmochimica Acta 73 (2009) 6631–6677

(Fig. 5B). These magnetites range in size from 10 to turally identical to those of the magnetite-rich rims, one dif- 150 nm, and are similar to those observed in the inner ference between them is that the veins contain few or no and outer magnetite rims (Figs. 5A and 5C). Although detectable Fe-sulﬁdes (Figs. 5A and 5D) arguing for a dif- magnetite-rich veins that cross-cut the outer cores are tex- ferent mode of formation or a diﬀerent generation. Author's personal copy

Origins of Magnetite in ALH84001 6643

Fig. 3D. Upper view: TEM image of ‘Texas’ FIB Section 2 extracted from the disk core (see 3A). Carbonate is coarse-grained (1–5 lm along the longest axis) and displays low porosity. Lower views: expanded view of ROI outlined by red box showing interlocking rhombohedral-shaped carbonate crystals visible at a tilt angle of +30°. Although few magnetites are apparent, this is misleading since at this angle few of the numerous magnetites present were in a strongly diﬀracting orientation. A rectangular-shaped void, 60 nm in length, is present in the upper right corner of the view and is shown enlarged on the right. This feature is consistent in both size and shape to negative crystals imaged by TEM (Viti and Frezzotti, 2001). (For interpretation of the references to color in this ﬁgure legend, the reader is referred to the web version of this paper.)

3.2.2. ‘Posterboy’ Section 2 and ‘Texas’ Sections 1 and 2 spersed both within these crystals, and along grain bound- The carbonate cores are predominantly composed of aries (Fig. 3C), are sub-micron ‘void’ spaces which blocky interlocking, irregularly shaped carbonate crystals demonstrate a range of morphologies from amoeboidal to (Figs. 3Cand 3D, 3F) ranging up to 5 lm in size. Inter- well faceted polygonal negative crystals (Fig. 3D). Only 3 Fig. 3C. Center: TEM image of ‘Texas’ FIB Section 1 extracted from the disk core (see 3A). The micron thick upper band of Pt was deposited as protection from Ga+ ion beam damage during FIB milling. The core region is composed of interlocking, mixed-cation, carbonate crystals ranging up to 5 lm in size cross-cut by multiple veins. This coarse-grained interior texture contrasts sharply with that of the fine-grained upper surface texture described earlier (see Fig. 1I). Two regions of interest (ROIs) are indicated by red and blue boxes. Top views: Expanded view of ROI outlined by red box showing several veins cross-cutting the host carbonate. Veins are composed of fine grained (<100 nm) randomly oriented carbonate grains intimately mixed with an amorphous S-bearing phase that appears homogeneously distributed. The associated EDX spectrum was acquired using 100 nm spot size and 1000 s dwell time. We suggest the veins formed by dissolution of core carbonate during exposure to a S-containing (possibly acidic) fluid. Lower views: expanded view of ROI outlined by blue box showing numerous nanophase magnetite and voids which appear to comprise <1 vol.% of the total FIB section in this TEM view (During full TEM rotation (±44°) a significant increase of magnetite crystals (up to 5 vol.%) is observed in this FIB section. This is due to image contrast obtained by the interaction of the electron beam with the sample. In a TEM image of an ALH84001 FIB section, denser areas appear darker due to scattering of the electrons in the sample. In addition, scattering from crystal planes introduces diffraction contrast. This contrast depends on the orientation of the magnetite crystal with respect to the direction of the incoming electron beam. Thus, tilting a crystal will cause the gray-level of that crystal to change. As a result, in a TEM image of a sample consisting of randomly oriented crystals, each one will have its own gray-level at a given orientation. In this way discrete magnetites can be distinguished from each other and their carbonate matrix at different tilt angles). Magnetites (e.g., arrows 1–2) and void spaces (e.g., arrows 3–4) are located at the margin of the false colored carbonate crystal. The concentration of magnetites and void spaces at grain boundaries may due to the migration of impurities which occurred during minor carbonate recrystallization. EDX spectrum of false color carbonate shows major Mg, Ca, and Fe with minor Si and Mn. The analysis spot size and acquisition time are 100 nm and 1000 s, respectively. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this paper.) Author's personal copy

6644 K.L. Thomas-Keprta et al. / Geochimica et Cosmochimica Acta 73 (2009) 6631–6677

Fig. 3E. Upper: TEM view of ‘Poster Boy’ FIB Section 1. This section was extracted from the rim zone (see Fig. 3B) and contains both inner and outer magnetite-rich rims separated by a magnesite band (as indicated in schematic inset). The underlying OPX displays saw-tooth projections, also known as microdenticular features, previously noted to be consisted with silicate dissolution (see Figs. 1F and 1G). ROI highlighted by the red box shows two subparallel, finger-like projections or veins, ranging from 2to5lm in length and 300 nm in width, which extend into the outer core carbonate. The top magnetite-rich vein originates from the uppermost surface (i.e., carbonate/Pt band interface) of the carbonate disk while the bottom vein originates from the inner magnetite-rich rim. Veins are texturally identical to the inner and outer magnetite-rich rims being composed of abundant nanophase magnetite embedded in a fine-grained carbonate matrix. Both rims and veins are enriched in Si, present as an amorphous phase (see Figs. 5A and 5D), compared to the surrounding carbonate matrix. However the veins are different from the rims in that they contain far fewer Fe-sulfides (see Fig. 5A). Although the general perception of the magnetite-rich rims is of simple well demarcated bands that are vertically contiguous, this is an over simplification as evidenced by the observation of these magnetite-rich veins. Magnetite abundances in veins mimics that of the rims (50–70 vol.%), while those in the carbonate cores and magnesite bands are much lower.

a small amount of the observed pore space (<10% of sur- single stoichiometrically pure Fe3O4 crystals (Fig. 5B). face area in the imaged sections) is associated with those By analogy with terrestrial carbonates, it is unlikely that grain boundaries (Fig. 3C) along which are also located this texture is primordial, but rather is the result of Author's personal copy

Origins of Magnetite in ALH84001 6645

Fig. 3F. Upper: TEM view of ‘Poster Boy’ FIB Section 2 (also see 3B). Lower: expanded view of ROI outlined by red box showing discrete nanophase magnetites, some which appear to be associated with voids while others appear to be encased in carbonate. Core carbonate typically displays low porosity (<1 vol.%). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this paper.) aggrading neomorphism (Folk, 1965), in which diagentic previously. In a small number of cases we have interpreted carbonate recrystallization has resulted in an increase in distinctive polygonally shaped voids to represent negative crystal size without a concomitant change in the overall crystals (see Fig. 3D). Cutting through the core carbonate chemistry since fine scale cation zoning is still preserved. are numerous branching veins (200 nm to several lmin During recrystallization, crystallographic imperfections length) that have fine grained porous texture (Fig. 3C). In such as void spaces would migrate out to grain boundaries, contrast to the veins previously described in the outer consistent with our observation that the majority of void carbonate core which contain embedded magnetite, these spaces are located at grain margins (e.g., Fig. 3C). At the veins do not, being composed of fine grained carbonate carbonate disk surface the interlocking blocky texture is intermixed with minor S, although no crystalline sulfides replaced by a fine-grain matrix as seen in Fig. 1I described or sulfates were observed.

Author's personal copy

Origins of Magnetite in ALH84001 6647 3 Fig. 3G. Center: TEM view of ‘Poster Boy’ FIB Section 4 showing complete rim structure (as illustrated in schematic inset; also see 3B). Outer core carbonate contains an extensive network of magnetite-rich veins, some which lie at the carbonate-OPX interface (e.g., base of the carbonate disk). Magnetite veins are texturally identical to the magnetite-rich rims and contain 50–70 vol.% magnetite embedded in a fine- grained carbonate matrix. The hole in the center of the section was produced by ion beam milling. In order to preserve the integrity of this FIB section, a region of the magnesite band (center of section) was not thinned to electron transparency (100 nm thick).This section is unusual in that, in addition to the archetypal carbonate rim structure with inner and outer magnetite bands separated by a magnesite band, it contains a third partial magnetite-rich rim that overlies the upper center of the magnesite band but does not extend down to the underlying OPX (see right schematic inset). Magnetite-rich veins cut across the outer core carbonate, extending from the upper carbonate surface to the carbonate- OPX interface. OPX is also present partially overlying part of the upper surface of the outer core (see carbonate schematic on left). Two ROIs in the magnesite band are indicated by red boxes. Upper: the ROI shown in the upper box demonstrates the presence of magnetite grains (arrows) embedded within the magnesite band, which are not obviously associated with a magnetite rich vein. The composition of one of these magnetites (red circle and arrow) and the surrounding magnesite matrix (blue circle and arrow) is shown in the EDX spectra to the right along with the resulting difference spectrum (green). Both spectra were obtained using identical spot sizes and dwell times (300 s). These EDX spectra demonstrate that the magnesite host matrix is essentially Fe-free while the embedded magnetite is conversely Mg-free. In such a case it is impossible, either thermodynamically and/or kinetically, for this magnetite to have arisen from partial thermal decomposition of the surrounding matrix. Lower: The ROI shown in the lower red box demonstrates the presence of magnetites (arrows) at the base of the magnesite band and lying parallel to the magnesite-OPX interface, which are also not obviously related to the presence of a magnetite-rich vein. As with the upper ROI, EDX spectra of one of these magnetites (red) and the surrounding magnesite (blue) in which it is embedded are shown to the right along with the difference spectrum (green). Both magnetite and magnesite spectra were obtained under the same analytical conditions. Again, the magnesite is Fe-free while the magnetite is conversely essentially Mg-free, the small Mg peak in the difference spectrum being due to magnesite either under or overlying the magnetite grain, since its size is comparable to that of the section thickness. The same argument developed for the upper ROI is also applicable, namely such a magnetite cannot have originated from partial thermal decomposition of the surround magnesite. As an illustration, the EDX analysis beam spot was refocused onto the upper left section of the ROI and the beam density increased to above the damage threshold of the magnesite to begin thermal decomposition. This resulted in the formation of an elliptical hole, but did not in the formation of any detectable magnetite. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this paper.)

In TEM thin sections while some magnetites appear carbonate phase using 3M ethanoic acid (CH3CO2H). completely encapsulated in carbonate, others are associated Compositional analyses by EDX showed that the vast with void space. The latter observation has been interpreted majority of these magnetites (>95%) were stoichiometri- 7 as evidence for their formation by thermal decomposition cally pure Fe3O4 within detection limits (Thomas-Keprta of the host carbonate (Barber and Scott, 2002; Brearley, et al., 2000a). While this approach yields an abundant pop- 2003), since the decarboxylation of siderite to magnetite re- ulation of magnetites for analysis, all information on their sults in a volume decrease of 50.5%, due to the higher distribution within the host carbonate was lost. In order 3 density of the product magnetite (qmagnetite 5.21 g cm ) to preserve this information in this study we performed 3 relative to reactant siderite (qsiderite 3.87 g cm ). How- in situ EDX analyses of individual magnetites present in ever, in many cases the void space volume observed is far the ‘Posterboy’ and ‘Texas’ thin sections. Our results indi- greater than that which could be accounted for density dif- cate that these magnetites share a similar physical size ferences. An alternative explanation is that during carbon- and morphology to those previous characterized and also ate recrystallization grain boundary migration allowed for appear to be stoichiometrically pure Fe3O4. In cases where initially unrelated impurities including void space and minor amounts of Mg, Si, and Ca were also observed they magnetites to relocate along grain boundaries, as is ob- were at, or below, levels that could be fully accounted for served in terrestrial carbonates (Kennedy and White, by spurious X-ray fluorescence8 from the surrounding 2001; Reid and Macintyre, 1998). carbonate. There have been several reports of individual ALH84001 magnetites having either an epitaxial or topotactic relationship to their host carbonate (Barber and Scott, 2002; Bradley 7 Based on thin film and doped glass standards the minimum et al., 1996; Brearley, 2003). Although this has again been level of detection (i.e., >3d background) for Mg and Mn in 100 nm interpreted as evidence for the partial thermal decomposi- magnetite crystals, with an integration time of 1500 s, is on the tion of the carbonate matrix we note that such relationships order of 0.05 wt.% for Mg and 0.02 wt.% for Mn. also result as a consequence of low temperature carbonate 8 This can occur through either primary or secondary X-ray recrystallization. Numerous observations at low tempera- fluorescence of surrounding carbonate. In the former case, this ture of epitaxial/topotactic overgrowths include quartz on arises through fluorescence of carbonate that either over or underlies magnetite in the section or through the scattering of the microfibrous opal-CT (SiO2 nH2O) (Cady et al., 1996)in marine environments, in vivo formation of calcium oxalate incident electron beam by the magnetite crystal into the surround- (CaC O ) on calcite (Geider et al., 1996), and growth of ara- ing carbonate. In the latter case, Fe Ka X-rays from the magnetite 2 4 can excite secondary X-ray fluorescence from the surrounding gonite (CaCO ) and calcite on strontium carbonate (SrCO ) 3 3 carbonate. Therefore the presence of trace to minor amounts of from solution (McCauley and Roy, 1974). Mg, Si, and Ca should not be construed to indicate their necessary inclusion as impurities within magnetite, a supposition that is 3.3.2. Stoichiometrical ‘Pure’ ALH84001 magnetite supported by the observation of either the greatly attenuated or the In previous studies, we characterized magnetites isolated complete absence of Mg, Si, and/or Ca in the subset of magnetites from entire carbonate disks by chemical dissolution of the that serendipitously bordered or project into void space. Author's personal copy

6648 K.L. Thomas-Keprta et al. / Geochimica et Cosmochimica Acta 73 (2009) 6631–6677

Fig. 3H. Upper: TEM view of ‘Poster Boy’ FIB Section 3 (see Fig. 3B). Magnetite-rich veins extend from the inner magnetite-rich rim into core carbonate. A magnetite-rich vein lies along the interface of the carbonate disk and the underlying OPX. Lower: expanded view of ROIs outlined by the red and blue boxes showing magnetite-rich veins ranging from 2to4lm in length and 250 nm in width, which extend into the outer core carbonate. Veins are texturally identical to the inner and outer magnetite-rich rims being composed of abundant nanophase magnetite embedded in a fine-grained carbonate matrix. Both rims and veins are enriched in Si, present as an amorphous phase (see Figs. 5A and 5D), compared to the surrounding carbonate matrix. However the veins are different from the rims in that they contain far fewer Fe- sulfides (see Figs. 5A and 5D). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this paper.)

3.3.3. Impure ALH84001 magnetite the Al and/or Cr may have been present as an unobserved Although the vast majority of the magnetites observed Fe-bearing phyllosilicate surface coating. In the present in situ in the ‘Posterboy’ and ‘Texas’ thin sections were study although we attempted to identify such a magnetite chemically pure, there were several notable exceptions. In surface phase we were unable to ﬁnd any evidence to sup- several cases, Cr at trace to minor levels was observed in port it on any of the magnetites analyzed in situ within a particular magnetite crystal, but absent in the surround- the carbonate. ing carbonate (Fig. 5C). This is similar to our previous While both Al and Cr are known to be readily incorpo- observation of Al and Cr in some magnetites liberated by rated into the inverse Fd3m magnetite spinel structure (Karl acid dissolution of ALH84001 carbonate (Thomas-Keprta and Deborah, 2004; Razjigaeva and Naumova, 1992; Tho- et al., 2000a). In that instance, Treiman (2003) argued that mas-Keprta et al., 2000a), neither has ever been docu- Author's personal copy

Origins of Magnetite in ALH84001 6649

Fig. 3I. Upper: TEM view of ‘Poster Boy’ FIB Section 5 (see Fig. 3B). Magnetite-rich veins extend from the inner magnetite-rich rim into core carbonate. ROI highlights a region of the magnesite band. Note: see Fig. 5A for quantitative element maps of another ROI of this FIB section. Lower: expanded view of the ROI outlined by the red box with corresponding quantitative Fe and Mg maps (Ka line). Vertical color bars to the right of each of the element maps represent the number of X-ray counts per pixel. The magnesite displays a blocky texture intimately associated with a minor Fe component present as a fine-grained, amorphous phase consistent with ferrihydrite. Influx of secondary (i.e., ferrous-ferric fluid(s)) after carbonate disk formation is likely responsible for the presence of this amorphous Fe phase. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this paper.) mented to substitute into the R3c calcite-type carbonate grain carbonate, or an intimate mixture of both. While se- structure. In the absence of any speculative surface coating lected area electron diffraction (SAED) patterns of both to explain their presence, those magnetite containing Al coarse and fine grain fractions are consistent with bulk and/or Cr could not have formed through thermal decom- FeCO3 (Fig. 6A; Table 4), they are compositionally distinct position of the surrounding carbonate. To suggest other- with respect to the degree of cation substitution of Fe by wise would require the presence of either a non-existent Mg. The fine grained carbonate contains only minor Al/Cr substituted precursor siderite, or the coexistence of amounts of Mg compared to the coarse grained carbonate an unknown and indeterminable Al/Cr phase miscible with which contains major amounts of Mg (Fig. 6A). In both the precursor carbonate that underwent simultaneous size fractions, however, Mn and Si (in an unidentified decomposition. amorphous phase) were observed uniformly as minor and trace components respectively (Fig. 6A). No evidence of 3.4. Roxbury siderite any Fe-oxide phases was noted.

3.4.1. Characterization of unheated Roxbury siderite 3.4.2. ‘Slow’ heated Roxbury siderite Unheated Roxbury siderite was observed to be com- ‘Slow’ heating of Roxbury siderite resulted in the for- posed of both coarse and fine-grain components mation of an optically black product composed of irregu- (Fig. 6A). The coarse fraction is characterized by lath- lar-shaped discrete crystals of magnetite, ranging from shaped grains ranging in length from P0.1 to 1 lm, 20 to 500 nm in the longest dimension (Fig. 6B; Table although the upper size limit is likely artificially constrained 5). Minor to trace amounts of both Mg and Mn were de- due to chattering during microtome cutting. In contrast, the tected in every magnetite crystal analyzed (Fig. 6C). SAED fine fraction is characterized by equant particles <50 nm in analysis (Fig. 6B; Table 5). Composition of the ‘slow’ diameter ranging in geometries from angular to pseudo- heated product phase (Fig. 6C) is consistent with a spherical (Fig. 6A). Mixing between the two size fractions (Mg,Mn)-ferrite ((Mg1xMnx)Fe2O4) product whose end was highly variable and sample dependant, giving rise to re- members are magnetite (Fe3O4), magnesioferrite gions either composed of almost entirely coarse or fine (MgFe2O4) and jacobsite (MnFe2O4)(Table 5). No Author's personal copy

6650 K.L. Thomas-Keprta et al. / Geochimica et Cosmochimica Acta 73 (2009) 6631–6677

Fig. 4. Upper: schematic representation of the typical carbonate rim structure based on the four ‘Poster Boy’ FIB rim extractions (‘Poster Boy’ Sections 1; 3–5). The inner and outer magnetite-rich rims are not simple radial bands but show a complex network of veins that penetrate both the magnesite rim and the outer core carbonate. These veins are magnetite rich and appear intimately associated with fine-grained Fe- sulfides phases and minor amorphous silica phase. In addition to those magnetites associated with veins, numerous individual magnetites are also distributed throughout the magnesite rims and outer core. As with the vein magnetites these appear chemically pure, that is they are free of cations that characterize the surrounding carbonate, but they are not associated with any sulfides or silica phases (see text for a detailed discussion). Lower: schematic representation of the typical carbonate core structure based on three core FIB extractions, two from ‘Texas’ carbonate and one from ‘Poster Boy’ carbonate (Section 2). In general, the inner core carbonate is composed of multiple interlocked blocky irregularly shaped crystals that are not crystallographically aligned. Rimming the boundaries between these crystals are individual magnetite grains that can account for up to several percent of the volume of the inner core carbonate. Unlike the magnesite band and outer core no magnetite rich veins were observed in inner core sections, however numerous porous fined grained carbonate veins intimately associated with a poorly defined S-rich phase that appears either amorphous or proto-crystalline are observed. These veins run in all orientations and do not correlate with crystallographic boundaries (also see Fig. 3C). Author's personal copy

Origins of Magnetite in ALH84001 6651

Fig. 5A. Upper left: ‘Poster Boy’ FIB Section 5 showing the location of a magnetite-rich vein that extends from the inner magnetite-rich rim into outer core carbonate. Expanded view of the ROI outlined by the red box on the upper right shows the intersection of the magnetite-rich vein with the inner magnetite-rich rim. Lower views: quantitative Si, Fe, Mg, and S maps (Ka line) of the ROI above. Vertical color bars to the right of each of the element maps represent the number of X-ray counts per pixel. The core carbonate in the lower left quadrant of the ROI, contains an approximately equimolar (Fe,Mg)-carbonate with an intimately associated minor Si component that is present either as an amorphous (glass, gel, colloid) or opaline phase. The magnetite-rich rim, upper right quadrant of the ROI, and magnetite-rich vein that penetrates the outer core carbonate also show a minor Si component but this is enriched by a factor of 3 relative to outer core carbonate. Additionally the rim shows a significant S enrichment due to the presence of nanophase Fe-sulfides, which contrast sharply with both the vein and outer core carbonate. We suggest that the presence of an amorphous Si phase in the magnetite-rich rim and vein suggests that the carbonate disk was penetrated, post formation, by a Si-rich fluid. Furthermore, since both the magnetite-rich rim and vein are texturally indistinguishable, the paucity of Fe-sulfides within the vein implies different origins for these two populations. evidence was observed for residual carbonate indicating 3.4.3. ‘Fast’ heated Roxbury siderite the decomposition reaction had proceeded to completion, ‘Fast’ laser heating of Roxbury siderite under vacuum nor was there any evidence for non-ferrous oxides such as resulted in a pronounced optical darkening of the periclase (MgO). original iridescent siderite surface (Fig. 6D). Microtome Author's personal copy

6652 K.L. Thomas-Keprta et al. / Geochimica et Cosmochimica Acta 73 (2009) 6631–6677

Fig. 5B. Paired EDX spectra taken in situ with a 50 nm beam spot size from ‘Poster Boy’ FIB sections illustrating the compositional inhomogeneity between individual magnetite grains and the respective host carbonate matrix in which they are embedded. Spectra on the left are of the individual magnetite grains while those on the right are taken from the carbonate matrix immediately (<100 nm) adjacent to the given magnetite. Both ALH 9-20 and ALH 9-01 paired spectra were acquired with a 1500 s dwell time, while the ALH 8-31 paired spectra were acquired with a 3000 s dwell time owing to a reduced count rate. In all three cases, major Mg and Mn present in the carbonate host matrix are, within experimental error, absent from the respective magnetite. In the case of ALH 8-31(0) spectra although trace amounts (<0.1 wt.%) of Mg are apparent in the magnetite spectrum, this most likely arises from spurious primary and secondary X-ray fluorescence of carbonate due to the thickness of the FIB sections that could be over- or underlying the magnetite grain. cross-sections through the altered surface indicated that the second region (Region 2) that extends downward a further laser induced heating penetrated down to a depth of 800 nm in which the carbonate appears to have under- 1.0–1.5 lm (see Fig. 6E). Based on textural differences, gone extensive vesiculation but has not otherwise decom- the ‘fast’ heated Roxbury siderite can be subdivided into posed as indicated by the absence of Fe-oxides (Fig. 6G; three regions (Fig. 6E). Beginning at the surface and Region 2, Table 4). Finally, below this the carbonate extending down to a depth of 400 nm (Region 1) the car- appears indistinguishable from the unheated starting bonate decomposition has gone to completion to produce material (Figs. 6A, 6E, Region 3, Table 4). closely packed assemblage of Fe-oxides ranging from 20 to 100 nm in size (Figs. 6E and 6F). In some cases, the 4. DISCUSSION temperature excursion created by the laser was sufficient to allowed surface-energy minimization of this Fe-oxide Partial thermal decomposition of sideritic carbonate has phase resulting in pseudo-spherical morphologies. As with been proposed as the origin of the nanocrystalline magne- the ‘slow’ heated siderite these Fe-oxides contain variable tite present within and throughout ALH84001 carbonate amounts of Mg but a uniform amount of Mn (Fig. 6F), disks (Brearley, 2003; Treiman, 2003). This reaction is most and SAED patterns implies the formation of both well commonly formulated as: and poorly crystallized (Mg,Mn)-ferrites (Figs. 6E and 6F; Region 1, Table 5) as defined above. Below this is a 3FeCO3 ! Fe3O4 þ 2CO2 þ CO ð1Þ Author's personal copy

Origins of Magnetite in ALH84001 6653

Fig. 5C. Upper left: ‘Poster Boy’ FIB Section 4 showing the location of a Cr-bearing magnetite. Expanded view of the ROI outlined by the red box (upper right) shows the Cr-bearing magnetite embedded in the inner magnetite-rich rim. Lower views: quantitative elements maps (Ka line) for Si, Fe, Cr, and O of the ROI. The rectangular-shaped Cr-bearing magnetite crystal is 200 nm in length embedded in the inner magnetite-rich rim enriched in Si. No Cr-bearing phases (i.e., phyllosilicates or iron-oxides/hydroxides) surrounded or coated this magnetite. The presence of Cr-bearing magnetite has important implications for thermal decomposition models proposed for the formation of ALH84001 magnetite since this magnetite cannot be the product of thermal decomposition of Cr-free carbonate. We suggest this magnetite is an allochthonous (i.e., detritial) component which became embedded in carbonate during or subsequent to disk formation. ALH84001 magnetites extracted from the carbonate matrix with similar compositions were described previously by Thomas-Keprta et al. (2000a). (For interpretation of the references to color in this ﬁgure legend, the reader is referred to the web version of this paper.)

Would such a process be feasible given the physical and ALH84001 is known to have experienced after carbonate chemical properties of ALH84001 carbonate? Since such a formation and which could have facilitated carbonate reaction could not have occurred in isolation, we also need decomposition. We argue this provides a natural and to address whether putative thermal decomposition is con- convenient way to categorize the various thermal decompo- sistent with the experimental observations of ALH84001 sition hypothesis into two groups – those that are subject to carbonates. Our approach to these issues was to ﬁrst con- kinetic control as exempliﬁed by the model of Brearley sider the nature and timing of the heating events that (2003) or those that are subject to thermodynamic control Author's personal copy

6654 K.L. Thomas-Keprta et al. / Geochimica et Cosmochimica Acta 73 (2009) 6631–6677

Fig. 5D. Upper left: ‘Poster Boy’ FIB Section 5. Expanded view of the ROI outlined by the blue box shows a magnetite-rich vein cross-cutting outer core carbonate. This vein is not associated with the magnetite-rich rim; it begins at the interface of outer core carbonate and an isolated patch of OPX adhering to the uppermost carbonate surface. Lower views: elements maps (Ka-line) for Si, Fe, O, S, Mg, and Mn of the ROI shown in the red box (upper right). The vein is texturally identical to the magnetite-rich rims being composed of nanophase magnetite embedded in a fine-grained carbonate matrix, but contains few Fe-sulfides and is enriched in Si by approximately a factor of three relative to core carbonate. Note Mg is not enriched at the interface of the vein/host carbonate indicating a lack of evidence for cation diffusion (see Section 4). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this paper.) as exemplified by the model of Treiman (2003). We consider then the nature and distribution of the product magnetites both the Brearley (2003) and Treiman (2003) hypotheses will be a reflection of both the initial carbonate composition first on their intrinsic merits as presented, and second in and the nature of the heating event from which they the light of new observations reported here. formed. At this stage in is important to make some general 4.1. The ‘ALH84001 Heating Event’ remarks on the mechanisms by which reactions in the solid-state are affected. Thermal decomposition of siderite If the partial thermal decomposition of sideritic carbon- represents the solid-state transformation of an ionic crystal ate is the source of magnetite in ALH4001 disk carbonates, lattice from rhombohedral ðR3cÞ to cubic (Fd3m) symmetry. Author's personal copy

Origins of Magnetite in ALH84001 6655

Fig. 6A. TEM images and accompanying EDX spectra of ultramicrotome sections (100 nm thick) of Roxbury siderite (bulk composition (Fe0.84Mg0.10Mn0.04Ca0.02)CO3 (Lane and Christensen, 1997)), composed of both coarse (lm) (upper view) and ﬁne (<100 nm) grained (middle view) components. From the EDX spectra it is apparent that while Mn appears to be uniformly distributed in both fractions, Mg is enriched by a factor of 4 in the coarse grained fraction. Both EDX spectra contain minor Si lack d-spacings consistent with crystalline silicates, indicating the presence of an intimately mixed, homogeneous, amorphous silica component (e.g., a glass, gel, or opaline phase). Both EDX spectra were acquired with an analysis spot size of 200 nm and acquisition time of 1000 s. Lower view: SAED pattern of Roxbury siderite (both coarse and ﬁne grained components). Observed d-spacings are consistent with a sideritic carbonate that is free of any Fe-oxides.

In general solid-state transformations are initiated by a existing between reactant and product. Nuclei growth can bond redistribution process, in the case of carbonates this be envisioned as the progressive advancement of this acti- is the reversible endothermic dissociation of the carbonate vated reaction interface into the surrounding reactant 2 2 anion CO3 O þ CO2 (de La Croix et al., 1998), fol- phase. Growth will continue until either the interface be- lowed by reorganization and remobilization of the isolated tween the product and reactant is terminated and/or the anion/cation aggregates thus formed into nascent nucle- thermal energy driving the decomposition is dissipated. ation sites. At the nucleation sites subsequent growth then This process of crystal nucleation and growth is a funda- occurs preferentially due to the enhanced reactivity of the mental assumption in solid state transformations of ionic reactant interface surrounding the nucleating product. This solids (Galwey and Brown, 1999). enhanced reactivity is derived in part from the lattice strain One of the deﬁning characteristics of magnetites induced by diﬀering unit cell parameters and symmetries embedded in ALH84001 carbonate is their sub-micron size Author's personal copy

6656 K.L. Thomas-Keprta et al. / Geochimica et Cosmochimica Acta 73 (2009) 6631–6677

Table 4 Carbonate d-spacings: unheated and ‘fast’ heated Roxbury siderite. Sideritea Magnesitea Rhodochrositea Mg–Fe carbonatea Roxbury siderite Roxbury siderite (unheated and ‘Fast’ (‘Fast’ heating-Region 2) heating-Region 3) This study This study hkl d-spacings d-spacings d-spacings d-spacings d-spacings d-spacings (nm) (nm) (nm) (nm) (nm) (nm) 012 0.359 0.354 0.366 0.354 0.36 0.36 104 0.279 0.274 0.284 0.275 0.28 0.28 006 0.256 0.250 0.262 0.251 0.26 110 0.235 0.231 0.239 0.232 0.24 0.23 113 0.213 0.210 0.217 0.210 0.21 0.21 202 0.196 0.194 0.200 0.194 0.20 0.20 024 0.179 0.184 0.183 – 0.18 – 108 0.173 0.177 0.177 0.173 – 0.17 a International Centre for Diﬀraction Data Powder Diﬀraction Database – PDF-2 release 1998.

distribution. If we assume that the rates of nucleation and 1 1 t 3:11 1011 ð2Þ growth can be approximated by a Arrhenius-type tempera- cooling 3 3 T 1 T 0 ture dependence (see Electronic Annex, EA-1) this would infer that the thermal decomposition of siderite occurred From this we estimate that the time for ALH84001 to cool at high temperature so as to facilitate rapid nucleation over to ambient (233 K; see EA-2) following its ejection would crystal growth. Furthermore it is implicit that the heating have been only a matter of hours, corresponding to cooling rate was fast enough to prevent significant carbonate rates of 1011–1012 K Ma1. decomposition from occurring before the peak temperature was reached. Otherwise, fewer and larger decomposition 4.1.2. Subsurface cooling products would be expected. Hence, if the magnetites in If cooling following a shock heating event had occurred ALH84001 formed by thermal decomposition, the responsi- while buried in the Martian regolith then the cooling rate ble heating event would have be characterized by a rapid would be a function of both the burial depth and the pro- thermal step to high temperature. Given these require- cess by which cooling occurs (e.g., hydrothermal cooling ments, and the evidence for multiple shock events experi- or thermal diffusion). The paucity of hydrous minerals in enced by ALH84001, the most probable geological ALH84001 (Brearley, 1998) implies limited hydrothermal scenario for decomposition would have been during the dia- exposure suggesting that heat diffusion would have been batic propagation of an impact-generated shockwave, the dominant cooling mechanism. In terrestrial impact either during the impact that ejected the ALH84001 mete- structures cooling rates in the melt-rich portions are on orite, e.g., Brearley (2003), or during an earlier pre-ejection the order of 107 K Ma1, while deeper within the impact impact while ALH84001 remained buried in the Martian structure cooling rates can be much slower, typically 102– regolith, e.g., Treiman (2003). In both cases the shock heat- 103 K Ma1. Several estimations for the cooling histories ing would have been characterized by a near instantaneous of the ALH84001 carbonate disks have been made based rise in temperature followed by a slower monotonic cooling on cation diffusion studies. Fisler and Cygan (1998) and back to ambient. The key difference between the two sce- Kent et al. (2001) have independently determined solid state narios is the timescale over which cooling could have oc- cation diffusion rates of Mg2+ and Ca2+ in calcite and curred. In the first case, cooling would have been rapid magnesite from which they suggest a lower limit for the and occurred through radiative loss into the vacuum, while cooling rate of 102–103 K Ma1. If the cooling rate in the second case, cooling would have been far slower were any slower the fine-scale chemical zoning of the occurring instead primarily through conduction into the carbonates could not have been preserved (Fig. 7A). Most surrounding regolith. recently, Domeneghetti et al. (2007) have used the degree of fractionation of Fe2+ and Mg2+ between the two non- 4.1.1. Ejection cooling equivalent octahedral sites, M1 and M2, in OPX 2þ 2þ 2þ 2þ It is difficult to tightly constrain the thermal history of ðFeM2 þ MgM1 () FeM1 þ MgM2Þ to estimate the cooling ALH84001 immediately following its ejection; nevertheless rate of the last significant thermal event experienced by we can make a reasonable order-of-magnitude estimation the OPX in ALH84001. Their results infer a closure temper- by modeling the meteorite as an idealized black body that ature, Tc, of 802 ± 30 K as a lower limit for the last impact cools solely by radiative emission (see EA-2). Under such heating event, with a corresponding cooling rate of conditions, the time to cool from some initial temperature 0.1 K day1. This suggests that if partial thermal decom- T0 down to temperature T1 is given approximately by the position of ALH84001 carbonates occurred as a conse- expression: quence of impact heating during burial, it would have Author's personal copy

Origins of Magnetite in ALH84001 6657

suggests a burial depth of 90 m. This is broadly consistent with other estimations of burial depth using cosmic ray exposure data for ALH840019 and the other Martian meteorites (Shuster and Weiss, 2005) that suggest burial depths of at least a few meters for nearly all of their histories. Mikouchi et al. (2006) calculated a burial depth of less than four meters for several nakhlites, which is the original vol- canic cooling depth.

4.2. Thermal decomposition models

The two heating scenarios outlined above provide the geological context for the decomposition models developed independently by Brearley (2003) and Treiman (2003). The former is based on carbonate decomposition occurring under “extreme disequilibrium conditions” in which “kinetics are the dominant controlling factor” determining the chemical and physical nature of the magnetites that are formed. This, although never explicitly addressed, is most consistent with the impact event that ejected ALH84001 from Mars. In the latter model, carbonate decomposition occurs “at some depth beneath the Martian surface where the pressure was greater than the atmospheric pressure and the temperature declined slowly.” Under these conditions equilibrium thermodynamics determined the chemical and physical nature of the magnetite formed. In this model, carbonate decomposition occurred prior to ejection, during an earlier epoch, and the eventual ejection event failed to obscure evidence of this earlier thermal episode (as an aside: the Brear- ley (2003) model is essentially based on inductive reasoning, that is it proceeds from a set of speciﬁc observations and experiments to a set of general conclusions. In contrast the Treiman model is based on deductive reasoning, that is, the fundamental postulates of thermodynamics are used to extrapolate the outcome of a single speciﬁc event). It should be noted that these models are necessarily contradictory; that is they cannot both have occurred since application of one model negates the applicability of the other. Indeed as Brearley (2003) noted, “equilibrium phase diagrams have only limited relevance... the kinetics of the reaction, rather than the thermodynamics, control the phase and their compositions.” It is also necessary in the Fig. 6B. Upper: TEM view of an ultramicrotome section of subsequent discussion to bear in mind in each model the Roxbury siderite after heating to 848 K using the LSQT method of key question is not whether a siderite-magnesite-[calcite] Golden et al. (2006). The heating rate was 1 K min1 (i.e., ‘slow’ heating) with the peak temperature being maintained for 24 h (see carbonate can undergo preferential decomposition of the Methods section). The grain size ranges from 20 to 200 nm. sideritic component, since to some extent this may well be Lower: SAED pattern of heated Roxbury siderite shown above. true. Rather the question is can the sideritic component of Measured d-spacings indicate the siderite has converted completely such a mixed carbonate undergo decomposition to the com- to magnetite. plete exclusion of the magnesite–calcite components, since this is what would be necessary to explain the observations of ALH84001 carbonates. The lack of any evidence whatso- had to have occurred on a timescale of only a few decades ever for the partial or complete thermal decomposition of or less. In either case to prevent complete, as opposed to the magnesite and calcite-rich components of ALH84001 partial, decomposition of the carbonate globules, the heat- carbonates requires that any proposed thermal decomposi- ing event would have had to have occurred under higher tion mechanism must be tightly limited to the siderite-rich pressure conditions than Mars atmospheric. Assuming a 3 Mars crustal density of qMars 3000 kg m (McSween, 2 2002), and a surface gravity of GMars 3.72 m s , the 9 Although ALH84001 probably formed as a cumulate in an pressure P in bars, at a depth d in meters, can be approxi- intrusion several 10s of kilometers in depth (McSween and mated as P qMarsGMarsd d. Hence a pressure of 10 bars im- 105 9 Treiman, 1998), its shock history suggests that it was later plies a burial depth of 9 m, while a pressure of 100 bars excavated to much shallower depths. Author's personal copy

6658 K.L. Thomas-Keprta et al. / Geochimica et Cosmochimica Acta 73 (2009) 6631–6677

Table 5 Magnetite d-spacings: ‘Slow’ and ‘fast’ heated Roxbury siderite. Magnetitea Magnesio-ferritea Jacobsitea Mg–Mn-ferritea Roxbury Siderite Roxbury Siderite (Mg-ferrite) (Mn-ferrite) (‘Slow’ Heated) (‘Fast’ Heated-Region 1) This study This study hkl d-spacings (nm) d-spacings (nm) d-spacings (nm) d-spacings (nm) d-spacings (nm) d-spacings (nm) 111 0.484 0.484 0.492 0.491 0.49 0.49 200 0.420 – – – 0.42 – 220 0.297 0.296 0.300 0.298 0.30 0.30 311 0.253 0.253 0.256 0.254 0.25 0.25 222 0.242 0.241 0.245 0.243 0.24 – 400 0.210 0.210 0.212 0.211 0.21 0.21 331 0.192 0.192 – – – – 422 0.171 0.171 0.173 0.177 – 0.17 a International Centre for Diﬀraction Data Powder Diﬀraction Database – PDF-2 release 1998.

carbonate only. Consequently, we address the merit of each dolomite-ferroan, dolomite-ankerite series (Milodowski model separately, and consider only those aspects of each et al., 1989) [which] exemplify the decomposition behavior model that relates specifically to chemical and physical nat- of complex carbonate solutions,” is cited. In both studies ure of magnetites formed from carbonate decomposition. “...the decomposition of dolomite and ankerite is complex and involves distinct stages of CO2 evolution, indicating 4.2.1. Kinetically controlled carbonate decomposition – that decomposition is taking place in different stages.” Brearley (2003) model From this is argued that both studies “...demonstrate that The key aspects, as relevant to the discussion herein, of preferential decomposition of one of two carbonate compo- the hypothesis proposed by Brearley (2003) can be con- nents (i.e., FeCO3, MgCO3) can occur, leaving the third densed into three main points: (CaCO3) intact.” We argue, however, that the relevance of this example is 1. ALH84001 underwent a rapid heating and cooling event in fact limited and more applicable experimental results as a consequence of a shock event. need to be considered. In the decomposition studies of 2. The most sideritic component of the carbonate disks ankerite-dolomite solid solutions (FexMg(1x)Ca(CO3)2; “...decomposed preferentially and at a significantly low- 0 6 x 6 0.7) discussed above, decomposition was inter- er temperature than either the magnesite, calcite, or rho- preted as proceeding in a series of three consecutive stepped dochrosite component in solid solution.” reactions, with the first step involves the preferential 3. The timescale for this thermal event was short enough to decomposition of the FexMg(1x)CO3 ‘component’ in the prevent any re-equilibration of the decomposition prod- presence of the CaCO3 ‘component,’ that is: ucts or compositional zoning patterns remaining in the for (0 6 x < 2/3)... residual carbonate. x 3x Fe Mg CaðCO Þ ! MgFe O þ 1 MgO x 1x 3 2 2 2 4 2 Critical to this hypothesis is the supposition that under conditions characterized by a “high degree of disequilib- x þ CaCO3 þ 1 CO2 rium,” partial decomposition of sideritic carbonate would 2 x favor the formation of pure magnetite. The rational is that þ CO ð3Þ “under disequilibrium conditions, the earliest phase that 2 forms is the one which is kinetically the easiest to nucleate, for (2/3 6 x 6 7/10)... in this case, a simple binary oxide [i.e., magnetite] rather 3Fe Mg CaðCO Þ ! Mg Fe O þ 3CaCO than a more complex solid solution [i.e., magnetite-magne- x 1x 3 2 3ð1xÞ 3x 4 3 sioferrite].” Several problems, however, exist with the þ 2CO2 þ CO ð4Þ Brearley (2003) hypothesis, particularly with over-general- izations that are made on the basis of limited or ambiguous While it is true that this represents an example of a prefer- experimental data. ential decomposition of one component (Fe, Mg) of a solid solution carbonate in the presence of another (Ca), it in and 4.2.1.1. Preferential decomposition of siderite solid solutions. of itself does not address the more relevant question as to Brearley (2003) argues that a substantial body of experi- whether the sideritic component of a siderite–magnesite mental data exists to support the contention “that a com- solid solution would also preferentially decompose (under plex carbonate solid solution can undergo progressive disequilibrium conditions). The reason that calcitic compo- decomposition, such that one component in solid solution nent of ankerite-dolomite solid solution does not undergo breaks down without the entire phase decomposing.” By decomposition with the siderite–magnesite solid solution way of example, “the behavior of ferromanganoan is that neither siderite nor magnesite can form a com- dolomite (Iwafuchi et al., 1983) and carbonates in the plete solid solution series with calcite. In contrast, Author's personal copy

Origins of Magnetite in ALH84001 6659

Fig. 6C. Upper views: TEM image and accompanying EDX spectra (1000 s acquisition time) of a grouping of magnetite crystals produced by the ‘slow’ heating of Roxbury siderite. Three magnetite crystals labeled ‘A’, ‘B’, and ‘C’ (outlined in green, blue, and red) are highlighted. Magnetites ‘A’ and ‘C’ contain only minor Mg relative to magnetite ‘B’, while minor Mn is present in all three grains in approximately the same concentration. Lower views: element maps (Ka line) of Fe, Mg, Ca, and Mn for the entire grouping of magnetites shown in the upper TEM view (red box), illustrating the compositional heterogeneity of Mg in the magnetite product (color bar to right of map represents the number of counts per pixel). Since the observed cation variations mirror those of the unheated Roxbury siderite (see 6A) there is no significant cation diffusion occurring during the ‘slow’ heating thermal decomposition. That is, the cation composition of a given magnetite grain simply reflects the cation composition of the progenitor carbonate grain from which it formed. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this paper.) Author's personal copy

6660 K.L. Thomas-Keprta et al. / Geochimica et Cosmochimica Acta 73 (2009) 6631–6677

Fig. 6E. TEM view of a region of an ultramicrotome section extracted from the upper surface of the ‘fast’ heated Roxbury siderite sample shown in 6D. The section represents a vertical proﬁle of the irradiated carbonate, with the top of the image corresponding to the uppermost irradiated surface. Three distinct textural regions are apparent and correlate with increasing vertical distance from the irradiated surface. Region 1 represents the uppermost half micron of the carbonate surface that was directly exposed to laser irradiation. This region is composed exclusively of impure magnetite. Underlying this is a layer, several microns thick, composed of a heavily vesiculated carbonate, Region 2, that delineates the transition from the completely decomposed carbonate of Region 1 to the unheated carbonate represented by Region 3. No magnetites were observed in either of the underlying Regions 2 and 3.

Fig. 6D. Optical views of Roxbury siderite before (upper view), and after ‘fast’ heating (lower view), by laser irradiation (pulsed 10.6 lmCO2 laser) in a vacuum. The unheated siderite is composed While the kinetics of the thermal decomposition of sid- of interlocking mm to sub-mm rhombohedral blocks displaying a erite and its solid-solutions have been extensively studied variety of iridescent colors due to birefringence. After ‘fast’ heating (Dhupe and Gokarn, 1990; Jagtap et al., 1992; Kubas the color of the carbonate surface is reduced to a silver-gray. (Note: and Szalkowicz, 1971; Zakharov and Adonyi, 1986) great the upper right section of the surface present in the unheated carbonate is absent in the image after heating due to thermally care needs to be taken in interpreting the results of such induced cleavage during laser irradiation.) studies since there are many discrepancies between datasets. The reasons appear to be two fold: first, reaction kinetics are sensitive to the physical parameters of the sample, siderite–magnesite solid solutions do form a complete series e.g., degree of crystallinity, surface area, presence of impu- between end members. Hence, the decomposition of rities, particles size, and the degree of compaction or poros- ankerite-dolomite solid solution is an example of a spinodal ity. Second, non-isothermal decomposition traces alone can decomposition in which a phase transform/separation to typically be fitted, with good correlation coefficients, to any calcite and a siderite–magnesite solid solution occurs prior number of kinetic models. Consequently we consider here to nucleation and growth of the decomposition product the work of Gotor et al. (2000) who have utilized both lin- phases. In fact, in these studies the siderite–magnesite solid ear heating rate (TG) and constant rate thermal analysis solution that initially decomposes forms a mixed cation (CRTA) procedures to investigate the decomposition kinet- spinel which would appear to directly conflict with the ics of both a pure ‘synthetic’ siderite and a ‘natural’ argument for which this reaction was proposed to support, Mg-bearing siderite (Fe0.626Mg0.300Mn0.052Ca0.022)CO3, namely that of the exclusive decomposition of the sideritic under identical conditions. Their results indicated that the component of a siderite–magnesite solid solution to form thermal decomposition reaction of pure siderite followed pure magnetite. The data from these experiments show that an Avrami-Erofeev A2 random nucleation and growth pure magnetite does not result, rather, a mixed Fe–Mg model, while the natural Mg-bearing siderite instead fol- spinel is the product. lowed an F1 unimolecular decay model. Fig. 7B shows Author's personal copy

Origins of Magnetite in ALH84001 6661

Fig. 6F. TEM image of an ultramicrotome section from Region 1 (see 6E) of ‘fast’ heated Roxbury siderite. The accompanying SAED pattern to the right indicates complete conversion of carbonate to polycrystalline magnetite. Some magnetites are well-formed, single crystals, for example grain ‘A’, while others, such as grain ‘B’, exhibit a mottled texture consistent with a paracrystalline structure that is intermediate between amorphous and crystalline states. The accompanying EDX spectra show that both magnetites ‘A’ and ‘B’ contain minor impurities Mg and Mn that were present in the progenitor Roxbury siderite. Both spectra were acquired using the same analytical spot size and acquisition time of 1000 s.

the model decomposition curves, calculated from the results are far from equilibrium. If this is true then the all the car- of Gotor et al. (2000) for both pure and natural siderites bonate decomposition studies listed in Table 6 can be con- under heating rates of 0.52, 5.2, and 52 K min1. Several sidered pertinent to the origin of magnetite in ALH84001 important observations can be drawn from the experiments: carbonate. Results from these investigations all show that decomposition of magnesite–siderite solid solution carbon- For both pure and natural siderites at all heating rates ate under a variety of atmospheres and a range of heating the onset of decomposition occurs at temperatures that regimes invariably produces an impure Mg-substituted greatly overstep the thermodynamic stability limit pre- magnetite in which there is little to no evidence of the pref- dicted for both pure siderite and magnesite–siderite solid erential decomposition of the siderite component of solid solutions. solution. While Brearley (2003) suggests “that Fe-rich The kinetic effect on the stability of siderite as a conse- phases with little or no Mg can be formed as metastable quence of magnesium substitution is at least an order phases during the very earliest stages of decomposition,” of magnitude larger than would be predicted from ther- it is unclear how this is relevant to magnetite in modynamic equilibrium arguments alone. ALH84001 carbonate for two reasons: first, the magnetites Mg-bearing siderite does not decompose in two or more in ALH84001, while sub-micron in size, are still well be- separate stages in which preferential decomposition of yond the size associated with primary nuclei (as demon- the siderite component occurs first. strated by their well defined crystal faceting); and, second, since at least 50% of the volume in the magnetite rims of From the first observation it is clear that even at heating ALH84001 carbonate is accounted for by magnetite and rates that are far slower than those associated with shock if these magnetites were formed from thermal decomposi- heating, decomposition still occurs under conditions that tion, this would require nearly 70% of the carbonate Author's personal copy

6662 K.L. Thomas-Keprta et al. / Geochimica et Cosmochimica Acta 73 (2009) 6631–6677

Fig. 6G. TEM view of an ultramicrotome section of the interface between Regions 1 and 2 (see 6E) of the ‘fast’ heated Roxbury siderite indicated by the dashed line. The SAED pattern (upper right) of Region 2 (lower right in TEM view) indicates that it is composed of poorly- to-moderately crystalline carbonate with no associated magnetite. The EDX spectra of both regions show they contain nearly identical amounts of Mg and Mn indicating that the cation composition of the carbonate is preserved in the magnetite decomposition product. That is, the ‘fast’ thermal decomposition of Roxbury siderite does not facilitate the formation of chemical pure magnetite with Mg and Mn cations preferentially partitioning into residual carbonate (and/or oxide phase). The residual C present in the EDX spectrum of Region 1 is due to the epoxy mounting substrate and not from unreacted carbonate. Spectra were acquired using the same analytical spot size (200 nm) and an acquisition time of 1000 s.

initially present in the rims to have decomposed.10 A reac- 4.2.1.2. Interpretation of siderite–magnesite decomposition tion having gone to this extent cannot be construed as being studies. The experimental studies of Koziol (2001) are used in its initial stages. To be able to rapidly decompose nearly in Brearley (2003) as evidence for the preferential decompo- 70% of a siderite-magnetite carbonate in a near instanta- sition of siderite in a siderite-magnesite solid solution, neous, highly non-equilibrium, decomposition event and where it is stated that “pure magnetite was formed, rather expect the product to be essentially pure magnetite is highly than Mg-bearing spinel.” However, this view is not reflected implausible. The Fe/Mg ratio of the decomposition prod- by Koziol (2001) who concludes merely that “preliminary ucts would more likely be similar to the Fe/Mg ratio of compositional data indicate that magnetite [formed from the initial carbonate and show little, if any, chemical frac- thermal decomposition] may have a magnesioferrite tionation except lost of CO and CO2, especially given the (MgFe2O4) component.” The primary results in the partial results of pulse laser heating experiments on mixed compo- siderite–magnesite decomposition studies presented in Koz- sition Roxbury siderite, which are able to duplicate and iol (2001) and subsequently in Koziol and Brearley (2002) even exceed impact shock heating rates and yet always pro- were: duced impure magnetites (see Results section ‘Roxbury Sid- erite’; also see Table 6). Residual carbonates all exhibited unit cell contractions, specifically “unit cell volumes [of the residual carbonate] are reduced by 0.27–1.09% with progressively more contraction seen at the more sid[ertite]-rich starting compo- 10 Assuming carbonate decomposes via the reaction 3FeCO ? 3 sitions.” (Koziol and Brearley, 2002). Fe O +2CO + CO then the volume ratio of carbonate to 3 4 2 “[m]agnetite formed during these experiments also magnetite (Vration) when the extent of reaction is a (0 6 a 6 1) is given by V ¼ 3V carbð1aÞ, where V and V are the molar exhibited contraction of its unit cell, as evidenced by ratio aV magn carb magn volumes of carbonate and magnetite respectively. Assuming the change in d-spacing of the mt 311 X-ray reflec- 3 3 Vcarb 2.9 cm and Vmagn 4.5 cm and solving for a when tion.” Specifically, the {311} d-spacing reported for Vratio 1, gives the extent of reaction as 67%. the spinel product formed from partial decomposition Author's personal copy

Origins of Magnetite in ALH84001 6663

Fig. 7A. Length scale, x, over which Mg2+ heterogeneities in Fig. 7B. Decomposition kinetics of pure and Mg-substituted calcite (CaCO3) would be homogenized by thermal diffusion- siderite as a function of linear heating rates. The two curves adapted from Kent et al. (2001). The thermal event that drives the corresponding to a heating rate of 0.52 K min1 are the best fits to diffusion is assumed to be characterized by an instantaneous rise in the experimental data of Gotor et al. (2000) corresponding to an temperature to some peak value, followed by slow monotonic Avrami-Erofeev A2 random nucleation and growth model for pure cooling back to ambient. Diffusion lengths for peak temperatures siderite, and an F1 unimolecular decay model for the Mg- T0 ranging from 500 to 900 K are shown by the solid lines in the substituted siderite. The curves corresponding to heating rates of plot. For a given initial cooling rate denoted on the abscissa axis, 5.2 and 52 K min1 are extrapolated using these kinetic models the diffusion distance, x, can be determined by tracing a vertical and the pre-exponential factors (A) and activation energies (EAct) line from the abscissa up to the intercept with the solid line calculatedpffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi by Gotor etÀÁ al. (2000). For pure siderite corresponding to the given peak temperature. From this intercept a EAct 6 da ¼ loge½1 a A exp RT dt where A = 7.88127 10 horizontal line traced to the ordinate axis gives the diffusion s1 and E = 106 kJ mol1, and for natural siderite distance below which compositional heterogeneities would not be Act ÀÁ da 1 a A exp EAct dt where A = 2.52625 109 s1 and preserved. The explicit form of the relationship between diffusion ¼ð Þ RT 1 EAct = 192 kJ mol . Numerical solutions for these differential distance and cooling rate is shown on the plot, where DT 0 is the 2+ diffusivity of Mg in calcite at the peak temperature T0, where equations were determined using the Adams–Bashforth–Moulton EAct is the diffusion activation energy and R is the gas constant. In predictor-corrector method (Mathews and Fink, 2004). The effect the terrestrial geological environment typical cooling rates range of Mg substitution in siderite has a profound consequence on the from 500 K Ma1 to 0.1 K Ma1. Cation heterogeneities in decomposition temperature and is substantially larger than that ALH84001 carbonate have length scales of x 6 1 lm hence the predicted based solely on equilibrium thermodynamics. cooling rate for any putative thermal event responsible for partial carbonate decomposition must have been extremely rapid in a chemical equilibria.” The geological context for the hypoth- terrestrial geological context. esis assumes that while ALH84001 was buried at “some depth beneath the Martian surface”, it was subjected to 11 of (Fe0.1Mg0.9)CO3 was 0.25010 nm, while that for an impact shock (I3 event ) that resulted in it being heated (Fe0.8Mg0.2)CO3 was 0.25227 nm. “rapidly to super ambient conditions,” resulting in the thermal decomposition of the most sideritic carbonate accord- Neither observation provides compelling evidence for ing the reaction: the preferential decomposition of siderite in a solid solution 3FeCO3ðFeMg solid solutionÞ Fe3CO4 þ 2CO2 þ CO ð5Þ carbonate to form pure magnetite. In the first case, lacking any data pertaining to the fraction of carbonate that under- The ALH84001 rock subsequently cooled back slowly to went decomposition, the spinel product formed could have ambient temperature over “a relatively short time geologi- cally...less than centuries.” Under these conditions the size had any composition between Fe3O4 and Mg0.3Fe2.7O4 (see EA-3). While in the second case, the 0.25010 nm line redistribution of magnetites is determined by kinetic factors since the nearly instantaneous temperature increase associ- ported by Koziol (2001) for the d311 line of magnetite ated with passage of the shock wave results in carbonate formed from the decomposition of (Fe0.1Mg0.9)CO3 is below any reported value of any documented (Mg,Fe)-spinel, decomposition under conditions far from equilibrium. The chemical composition of these magnetites, is however, and most likely assignable to the d006 line of the undecom- posed starting carbonate (see EA-3). determined subsequently by thermodynamic constraints during cooling where re-equilibration of magnetite with 4.2.2. Thermodynamic carbonate decomposition – Treiman the host residual carbonate is presumed to have occurred. model The conditions of such re-equilibration implicitly requires Treiman (2003) has proposed a “comprehensive abiotic a closed system in which the 2CO2 + CO gas phase evolved hypothesis” to explain the presence of fine-grained magne- during the initial decomposition is not dissipated or diluted, tite embedded in ALH84001 carbonate disks that “invokes high pressures, significant time for chemical reactions and 11 I3 event in shock chronology proposed by Treiman (1998). Author's personal copy

6664 K.L. Thomas-Keprta et al. / Geochimica et Cosmochimica Acta 73 (2009) 6631–6677

Table 6 Previous decomposition studies of natural and synthetic siderites. Composition of Unheated Atmosphere Decomposition Temperature and References Carbonate Product Phases

Synthetic solid solution siderite O2 770 °C(Chai and Navrotsky, 1996a) (0.0 to 100.0 mol.% FeCO3) Solid solutions from 0–66.6 mol.% FeCO3 ) magnesioferrite spinel (MgFe2O4) and MgO Solid solutions from 66.7 to 100 mol.% FeCO3 ) magnesioferrite spinel (MgFe2O4) and Fe2O3 a Ankerite solid solution O2 770 °C(Chai and Navrotsky, 1996b) Ca(FexMg1x)(CO3)2 CaO where 0 1 MgO Ca2Fe2O5

(Fe0.7Mg0.3)CO3 Vacuum Decomposition begins 452 °C(Gotor et al., 2000) Mixed Mg–Fe oxides No MgO

Mixed cation FeCO3 N2 and O2 Decomposition begins at lower T in N2 (Gallagher and Warne, 1981b) (1) 95.3 mol.% FeCO3, 4.2 mol.% In N2: mixed Mg–Mn ferrites MnCO3, 0.4 mol.% MgCO3 In O2: mixed Mg–Mn ferrites (2) 85.5 mol.% FeCO3, 4.0 mol.% and hematite MnCO3, 7.6 mol.% MgCO3, Spinel impurities increase 0.9 mol.% CaCO3 with increasing carbonate (3) 81.7 mol.% FeCO3, 6.4 mol.% impurities MnCO3, 9.5 mol.% MgCO3, 1.5 mol.% CaCO3 Mn-bearing natural siderite Air 480–530–700 (°C) (Hakonardottir et al., 2002) Hematite (initial) Mn-ferrite (ﬁnal)

Siderite: 79.16 wt.% FeCO3, N2 and O2 0 to 1000 °C(Gallagher and Warne, 1981a) 1.04 wt.% MnCO3, 16.40 wt.% N2: major impurities incorporated MgCO3, 3.76 wt.% Fe2O3, into the spinel product 0.43 wt.% CaCO3 O2: no spinel intermediate phases detected

Siderites: Approximate mole fraction N2 Onset and Peak Temperature (Dubrawski, 1991b) Fe (Fe/(Fe + Mn + Mg + Ca)) of Decomposition (°C): (1) 0.0 (1) 560, 640 (2) 0.3 (2) 530; 650 (3) 0.6 (3) 480; 580 (4) 0.7 (4) 450; 540 (5) 0.8 (5) 440; 540 (6) 0.9 (6) 440; 540 Mixed composition products from mixed composition siderite Ferroan magnesite ) Mg-ferrite Mn–Mg–siderite ) Mn-Mg-Fe-oxides Bakal siderite (impure) Air Decomposition begins 350 °C(Bagin et al., 1974) MgFe2O4

Siderite: (Fe0.79Mn0.12Mg0.07Ca0.02)CO3 Air Magnesiowustite (Isambert and Valet, 2003) 700 °C Mn-magnetite

Ankerite: (Ca(Mg,Fe,Mn)(CO3)2 CO2 (1 atm) Siderite, 500–700 °C(Cohn, 2006) Ankerite, 600–800 °C Signiﬁcant amounts of Mn and Mg in ferrites Author's personal copy

Origins of Magnetite in ALH84001 6665

Table 6 (continued) Composition of Unheated Atmosphere Decomposition Temperature and References Carbonate Product Phases 3 Copper Lake Siderite Impact shock CO2 (10 >470 °C(Bell, 2007) 0 to 21 mol.% Mg torr) 30 to 49 GPa Ferrites with range in composition (0 to 21 mol.% Mg) Siderite: Laser under vacuum 8.4 Mn-“magnetite-like” (Isambert et al., 2006) Fe0.79Mn0.12Mg0.07Ca00.02 to 25.9 GPa,

(1a,b) Siderite N2 and O2 Decomposition begins at lower T in N2:(Ware and French, 1984) (2a,b) Magnesite (1a) N2 450–550 °C (3a,b) Calcite (1b) O2 increased by 50 °C over 1a (2a) N2500–600 °C (2b) O2 increased by 70 °C over 2a (3a) N2680–780 °C (3b) O2 increased by 150 °C over 3a

Siderite (>90 wt.% FeCO3)N2,O2,CO2,H2O Siderite decomposition T increases (Hurst et al., 1993) Ca-siderite (70–90 wt.% proportional to extent of

FeCO3, 10 wt.% CaCO3) substitution by Mg, Mn, and Ca Mn-siderite (up to 28 wt.% Decomposition T decreases with increasing FeCO3 MnCO3) Mg-siderite (10–30 wt.% Non-linear decrease in decomposition T with increase MgCO3) in partial pressure H O High Mg-siderite (30– 2 Decomposition T increases 60 wt.% MgCO ) 3 non-linearly with increasing CO Ferroan magnesite 2 Low partial pressure (<25 kPa) CO (>60 wt.% MgCO ) 2 3 produces increase in decomposition T (e.g., in

N2,FeCO3 decomposes at 375 °C; in CO2, FeCO3 decomposes at 500 °C) a Results typical for decomposition of carbonates in ankerite–dolomite solid solution series. Compositions of phases formed are variable, determined by compositions of the precursor carbonates. Formation of chemically pure Fe3O4 by thermal decomposition of carbonates in this series has not been reported.

and furthermore the fO2 is solely defined by this 2CO2 +CO tion involving the presence of a CO gas phase product are gas phase. Unfortunately, the merit of the Treiman (2003) wrong. Using the most recent version of the thermody- hypothesis is undermined by the use of erroneous or namic dataset used by THERMOCALC v3.26 (and hence for re- questionable thermodynamic data and the application of ferred to as HP9812), in combination with the Peng– incongruous experimental results as discussed below. Robertson–Gasem cubic EOS (Gasem et al., 2001), and assuming Lewis Randall fugacity mixing rules (Lewis and 4.2.2.1. Siderite stability field. The thermodynamic stability Randall, 1961), the correct siderite stability P–T plot is field of siderite is calculated incorrectly in Treiman (2003) shown in Fig. 7C. From this it is apparent that the correct and this subsequently affects many of the later calculations. thermodynamic decomposition temperature of siderite is In a closed system, the decomposition of siderite according actually considerably higher than that shown Figs. 3 and to reaction (5) will be a function of the intensive variables of 4 of Treiman (2003) (a more comprehensive description of pressure and temperature, and if the siderite is not pure, but the relevant thermodynamic calculations can be found in rather present as a solid solution, its activity as well. In cal- EA-4). culating the siderite stability field Treiman (2003) used the thermodynamic software package THERMOCALC (Holland 4.2.2.2. Composition of magnesioferrite–magnetite solid and Powell, 1998; Powell et al., 1998) in combination with solutions. The equilibrium extent of Mg2+ substitution in a regular solid solution activity-composition model (Chai the magnetite product produced from the decomposition and Navrotsky, 1996a) to describe the siderite–magnesite of sideritic carbonate can be addressed thermodynamically solid solution series. Regrettably, the results of these calcu- by casting an Mg2+-substituted magnetite as a solid lations are incorrect since it appears that no consideration solution between magnesioferrite and magnetite proper, was made to address the absence of an appropriate equa- that is: tion-of-state (EOS) for CO in the THERMOCALC package, which only includes a compensated Redlich-Kwong EOS to determine fugacities for CO2 and H2O. Hence, all the calculations in Treiman (2003) relating to siderite decomposi- 12 http://www.earthsci.unimelb.edu.au/tpg/thermocalc Author's personal copy

6666 K.L. Thomas-Keprta et al. / Geochimica et Cosmochimica Acta 73 (2009) 6631–6677

2Fe3O4 þ 3MgCO2 3MgFe2O4 þ 2CO2 þ CO ð7Þ

Additionally it is also used to deﬁne the fO2 conditions for the reduction of magnesioferrite to magnetite proper and periclase via the following reaction: 1 3MgFe O 2Fe O þ 3MgO þ O ð8Þ 2 4 3 4 2 2 Nevertheless, thermodynamically under the reaction conditions presumed for siderite decomposition (“pressures on the order of tens to hundreds of bars” and temperatures in the range of 600–800 K) CO would be unstable with respect to disproportionation via the Boudouard reaction:

2CO CO2 þ C ð9Þ Consequently, the thermodynamically favored reaction for Fig. 7C. Univariant equilibrium decomposition curves (solid lines) thermal decomposition of sideritic carbonates is: calculated for siderite-magnesite solid solutions in P–T space subject to the reaction. Calculations were made using thermody- 6FeCO3 2Fe3O4 þ 5CO2 þ C ð10Þ namic data from the updated Holland and Powell (1998) database Fig. 7D shows the P–T phase stability curves for both (THERMOCALC Ver. 3.26) and assuming a Peng–Robinson–Gasem EOS for gas phase species (Gasem et al., 2001). The numbers pure siderite and siderite–magnesite solid solutions calcu- associated with each curve represent the mole fraction of siderite lated using thermodynamic data from THERMOCALC and assuming the regular solid-solution mixing model de- present in the carbonate solid solution (i.e., x in (FexMg1x)CO3) with the curve labeled “1.0” representing pure siderite (x = 1). The scribed previously for siderite–magnesite solid solutions. set of four curves shown by the dashed lines represent those While Treiman (2003) does note that formation of a calculated erroneously by Treiman (2003). graphite phase ought to be one of the products of siderite decomposition under conditions approaching equilibrium, 6 6 1 it is argued that “formation of graphite, the thermody- MgxFe3xO4 ) x MgFe O4 þð1 xÞFe3O4 0 x 2 3 namically stable phase, would be kinetically hindered.” ð6Þ In other words, although the Boudouard reaction is thermodynamically spontaneous, graphite deposition in the Under the conditions of siderite decomposition pro- absence of a catalyst is minimal due to the high activation posed by Treiman (2003) two thermodynamic methods barrier. For practically all laboratory investigations of sid- are developed to provide limits to the mole fraction of erite decomposition, it is true that in most cases experi- magnesioferrite present in ALH84001 magnetites. Both ap- ments are conducted over time intervals of less than a proaches appear to yield similar results, namely that the few days. Hence, while French and Eugster (1965) noted mole fraction of magnesioferrite is expected to be very low and probably undetectable by TEM/EDX. This is, however, misleading as both approaches use either erroneous or logically inconsistent data and the similarity in the results for each is simply coincidental, as discussed in EA- 5, where we show that because of the omission of an equation of state in the THERMOCALC package used by Treiman (2003), the calculated minimum concentration of Mg-spinel component in the Fe-spinel decomposition product is 3 higher than estimated by Treiman (2003) and would be detected by TEM/EDX analysis. As shown in EA-5, the use of data from multiple sources in the THERMOCALC treatment of oxygen fugacity also produces a signiﬁcant underestimation of the activity of the magnesioferrite component in the resulting decomposition product. A more consistent database predicts that the magnesioferrite component would be much higher and detectable by TEM/EDX techniques.

4.2.2.3. Where is the graphite? Metastability of CO –CO gas Fig. 7D. Univariant decomposition curves for siderite-magnesite 2 solid solutions in P–T space according to the reaction. The mixture and the formation of graphite. Central to the numbers associated with each curve represent the mole fraction of Treiman (2003) hypothesis is that the decomposition of siderite present in the solid solution, with the curve labeled “1.0” siderite occurs in a closed system via reaction (1) to produce representing pure siderite. Calculations were made using thermo- magnetite and a 2CO2 + CO gas phase. It is the presence of dynamic data from the updated Holland and Powell (1998) 2+ this 2CO2 + CO gas that allows re-equilibration of Mg database (THERMOCALC Ver. 3.26) and assuming a Peng–Robin- between magnetite and carbonate via the following reaction: son–Gasem EOS (Gasem et al., 2001) for gas phase species. Author's personal copy

Origins of Magnetite in ALH84001 6667

that “it is possible that rates of equilibration between in P T fO2 space, which is defined by the bivarient reac- graphite and gas are slow enough so that a gas phase with tion13 surfaces representing oxidation to either hematite or metastably low CO2/CO ratios could be maintained with- magnetite, and the interconversion between hematite and out precipitation, particularly at low temperatures” their magnetite, that is: experimental data indicated “that equilibrium is attained 2Fe O þ 4CO 4FeCO þ O ð11Þ through precipitation of graphite during runs of long 2 3 2 3 2 (two weeks) duration.” Tamaura and Tabata (1990) 2Fe3O4 þ 6CO2 6FeCO3 þ O2 ð12Þ showed conversion efficiencies near 100% using cation ex- 6FeO3 2Fe3O4 þ 6C þ 5O2 ð13Þ cess magnetite (Fe2.887O4) as a catalyst in CO2 to form 6Fe2O3 4Fe3O4 þ 6C þ 5O2 ð14Þ graphite at 563 K. It is should also be noted that the From inspection of reactions (12) and (13) it is can be seen use of CO2–CO gas furnace mixtures to fix fO2 in mineral equilibria investigations are usually restricted to tempera- that their equilibrium surfaces in P T fO2 space will inter- tures above 1200 K due to problems associated with sect on the graphite buffer surface defined by the reaction: graphite precipitation (Huebner, 1975; Jurewicz et al., C þ O2 CO2 ð15Þ 1993). These studies all suggest that graphite should be one of the decomposition phases in ALH84001, if siderite To visualize these relationships it is convenient to consider 2-D slices through the relevant region of P T f space was decomposed to form magnetite. O2 Although no exact analogues to ALH84001 carbonate perpendicular to one of the axis vectors, in our case this will have been identified in the terrestrial geological record, be pressure. Under isobaric conditions the bivarient reac- ancient (>3.5 Ga) secondary carbonate deposits occurring tion surfaces for reactions (11–14) appear as univarient curves in T f space. Provided that the partial pressure in vein fractures of rocks forming the in 3.8 Ga old Isua O2 of O defined by the value of f is negligible in comparison supracrustal belt in southern West Greenland do provide 2 O2 an interesting comparison (Zuilen et al., 2002). These car- to the total gas pressure of the system, the position of the bonate deposits are intimately associated with the pres- univarient curves defined by reactions (11–14) can be calcu- ence of magnetite and graphite inclusions. Carbon lated by equating the respective thermodynamic equilibrium constants to the Gibbs free energy of reaction, that is: isotope analyses indicate the parentage of the graphite to be derived from the carbonate, and it is generally ac- DGP 1;T 1 f exp rxn 16 cepted that both the magnetite and graphite formed by O2 ¼ ð Þ R T 1 “the disproportionation of Fe+2-bearing carbonates at high temperature” as shown in reaction (10) (Zuilen et al., 2002). Since no graphite has ever been observed by us in any ALH84001 carbonate, if we assume the magnetite formed 13 The Gibbs phase rule x = g + m / r establishes a relation- by thermal decomposition then the time available for the ship between the variance (x) of a thermodynamic system and the attainment of thermodynamic equilibrium would have had number of components (g), the number of intensive variables (m), to have been short, perhaps on the order of weeks (or less the number of phases (/) and the number of independent chemical depending on the ability of magnetite to promote precipita- reactions that are occurring in the system (r). In the Fe–C–O tion of graphite through surface catalysis). This is a far more system there are three components (i.e., Fe, C, and O) so g =3, while in P T f space there are three intensive variables (i.e., rigorous constraint that the “...less than centuries” as pro- O2 P, T and f ) and so m = 3. Thus the Gibbs phase rule can be posed by Treiman (2003), and would arguably be too short O2 rewritten as x =6 / r. For reactions (11, 12) we have two to allow any significant degree of chemical equilibration be- solid phases in equilibrium with a gas so / = 3, and setting r =1we tween magnetite and carbonate in ALH8001. find the variance x = 2 defining a surface in P T fO2 space.

Under isobaric conditions in T fO2 space the number of intensive variables is reduced to m = 2 (i.e., T and f ) so that x = 1 defining 4.2.2.4. Siderite–magnetite stability and oxygen fugacity. In O2 the thermodynamic model of siderite decomposition it has a univariant curve. In the case of reaction (13) before applying the phase rule it must necessary to realize that we are implicitly been implicitly assumed that the system is closed. Although assuming the presence of an imaginary or inert gas is also present in not specifically a criticism of the Treiman (2003) hypothesis the system otherwise we would not be able to vary the partial it is useful at this point to consider the consequences if this pressure of O2 since it would be the only gas in the system and were not the case. Specifically let us consider the situation in hence would always be at the system pressure P. We account for which the fO2 is influenced by the external environment, as this imaginary or inert gas by increasing the number of components is often the case in geological settings. Since the thermal necessary to describe the system, i.e., g = 3 + 1 = 4. Since we have decomposition of siderite to magnetite involves a change three solid phases in equilibrium with a gas we have / = 4 and in oxidation state of the Fe cation we can expect there to r = 1. Now applying the phase rule we obtain a variance

x =4+3 4 1=2 in P T fO2 space and x =1 in T fO2 be only a limited range of fO2 values for which both siderite and magnetite can coexist. This should be contrasted with space. Finally for reaction (14) since C is absent the number of components is two (i.e., Fe and O) + the imaginary or inert gas the analogous decompositions of magnesite to periclase or necessary allow variation of the partial pressure of O , hence g =3. calcite to lime which are f invariant since they involve 2 O2 There are two solids and a gas phase described by a single reaction no changes in redox. In order to explicitly determine the so / = 3 and r = 1. Hence from the phase rule the variance fO dependence of the partial decomposition of siderite to 2 x =3+3 3 1=2 in P T fO2 space and x =1 in T fO2 magnetite we need to consider the stability ﬁeld of siderite space. Author's personal copy

6668 K.L. Thomas-Keprta et al. / Geochimica et Cosmochimica Acta 73 (2009) 6631–6677

Fig. 7E. Siderite stability field in P–T-space, using thermodynamic data from the updated Holland and Powell (1998) database (THERMOCALC Ver. 3.26) and assuming a Peng–Robinson–Gasem EOS for gas phase species (Gasem et al., 2001). Each plot represents a 2-D slice through P– T-space under isobaric conditions corresponding to pressures of 1, 10, and 100 bars, respectively. The siderite stability field is indicated by the magenta plus gold shaded wedge (shown outlined by the black dashed line in the middle plot) that splits the hematite and magnetite fields. The presence of siderite coexisting with magnetite in the absence of graphite occupies a restricted region in T-space defined by the curve, which defines the upper limit of the siderite stability field. This line connects the two invariant points in T-space corresponding at lower temperature to the coexistence of siderite + magnetite + hematite + gas, and at higher temperature to siderite + magnetite + graphite + gas. Note that the siderite stability field is itself divided into the magenta and gold shaded regions by the black line corresponding to the equilibrium

2CO C þ CO2. (For interpretation of the references to color in this ﬁgure legend, to reader is referred to the web version of this paper.)

P 1;T 1 where DGrxn is the free energy of the given reaction calcu- temperature limits for reaction (12) are bound by two lated at pressure P1 and temperature T1. In the case of reac- invariant points in T fO2 space. The low temperature limit tion (15) this is complicated by theÀÁ necessity of taking into corresponds to the intersection of reactions (11, 12, and 14) 1 account the Boudouard reaction C þ 2 O2 CO so that at which point siderite is in equilibrium with both hematite thermodynamic expression for the dependence becomes: and magnetite according to the equilibrium15: P 1;T 1 2 DGrxn Fe2O3 þ 4CO2 5FeCO3 þ O2 2Fe3O4 þ 6CO2 ð18Þ fO2 ¼½cCO P CO exp ð17Þ R T 1 While the higher temperature limit corresponds to the intersection of reactions (12, 13) and the graphite buffer reaction where cCO is the pure gas fugacity of CO at a partial pres- P 1;T 1 (15) at which point siderite is in equilibrium with magnetite sure PCO (see EA-6). The values of DG for each of the rxn 16 reactions can be evaluated using the same procedure as that and graphite according to the equilibrium : described for the siderite decomposition reaction (see EA- 4). Using thermodynamic data from the most recent HP98 dataset and assuming a Peng–Robertson–Gasem cubic EOS for gas phase species the calculated univarient 15 Applying the Gibbs phase rule we start with x = g + m p r. curves for reactions (11–15) at isobaric pressures of 1, 10 In the Fe–C–O, system we have three components plus an and 100 bars are shown in Fig. 7E. imaginary or inert gas (see footnote 14) so g = 3 + 1 = 4. There From Fig. 7E the stability field for siderite can be viewed are three solid phases in equilibrium with a gas phase and so / =4. as a narrow wedge separating the hematite and magnetite The system has two independent chemical reactions (11) and (12) and so r = 2 (note reaction (14) is not independent since reaction stability fields that expands upward to lower fO values with 2 ð13Þ¼1 reaction (11) – reaction (12)). Hence in T f space increasing pressure. Under the range of temperatures and 2 O2 (m = 2) the variance is given by x =4 2 4 + 2 = 0 defining an pressures considered (350–650 K and 1–100 bar) siderite is 14 invariant point. only thermodynamically stable at relatively low f values 16 O2 Again applying the Gibbs phase rule x = g + m / r,we with the upper section of the stability wedge defined by the have three components for the Fe–C–O system plus an imaginary interface between the siderite + gas magnetite + gas or inert gas so g = 3 + 1 = 4. There are three solid phases and one fields corresponding to reaction (12). The upper and lower gas phase so x = 4, and we have two independent chemical reactions (12) and (13) so r = 2 (note reaction (15) is not 1 independent since reaction ð15Þ¼6 (reaction (12) + reaction 14 f T f For reference O2 values for the current atmospheres of Earth (13)). Hence in O2 space (m = 2) the variance is given by and Mars are 2 101 and 1 104 bar, respectively. x =4+ 2 4 + 2 = 0 again defining an invariant point. Author's personal copy

Origins of Magnetite in ALH84001 6669

Table 7 ALH84001 carbonate compositions and inferred magnetite volumes. a Outer magnetite rim Vfraction magnetite

Initial carbonate Fe0.424±0.068Mg0.558± 0.072Ca0.065±0.015Mn0.012±0.005CO3 Residual carbonate Fe0.227±0.057Mg0.716± 0.024Ca0.047±0.035Mn0.010±0.003CO3 15.2% Inner magnetite rima

Initial-carbonate Fe0.752±0.144Mg0.241±0.143Ca0.004±0.005Mn0.003±0.001 CO3 Residual carbonate Fe0.393±0.032Mg0.599±0.031Ca0.007±0.001Mn0.002±0.002CO3 42.9% a From Treiman (2003).

2Fe3O4 þ 6CO2 6FeCO3 þ O2 2Fe3O4 þ 6C þ 6CO2 have had a higher siderite content and a correspondingly ð19Þ lower magnesite content than the remaining residual carbonate. If we let the composition of the initial carbonate Hence, it is therefore apparent that the partial decomposition be Fe Mg CO , and the composition of the residual x0 ð1x0Þ 3 of siderite to magnetite in the absence of graphite is only pos- carbonate be Fe Mg CO , then the relevant chemical x1 ð1x1Þ 3 sible under an extremely limited set of fO2 conditions reaction is: (44:46 < fO < 44:30 @ 1 bar; 38:15 < fO < 37:95 2 2 Fe Mg CO a Fe Mg b Fe O x0 ð1x0Þ 3 ! x1 ð1x1Þ þ 3 4 @ 10 bar; and, 31:95 < f O2 < 31:67 @ 100 bar). þ 2b CO þ b CO ð20Þ 2 4.2.2.5. Why is the residual carbonate composition so where a ¼ 1x0 and b ¼ 1 x0x1 . Several important different?. Table 1 of Treiman (2003) gives the composition 1x1 3 1x1 of individual carbonate grains present in both in the inner observations can be immediately deduced: and outer magnetite-bearing layers. The average composition of these carbonate grains in the two layers therefore de- The number of moles of Fe and Mg in both the initial fines the residual carbonate composition that would have carbonate, and the carbonate–magnetite product are had to have been in equilibrium with magnetite during equivalent. Hence, the composition of the initial carbon- decomposition under thermodynamic control. Table 7 ate can be directly determined from the bulk composi- shows the average carbonate composition of both inner tion of the carbonate–magnetite product. and outer magnetite layers based on data from Table 1 of The number of moles and hence volumes of the residual Treiman (2003). This suggests that in the outer magnetite carbonate and magnetite products are strictly bound by layer a residual carbonate with a 23 mol.% siderite compo- the mole fraction of Fe in the initial and residual carbon- nent would have been stable with respect to further decom- ates (i.e., x0 and x1). position. Yet, the residual carbonate in the inner magnetite The greater the proportion of magnetite product the rim has a siderite component of 39 mol.%, almost twice smaller the mole fraction of Fe in the residual carbonate that of the outer rim. Such a carbonate would not have since 1 > x0 > x1 >0. been thermodynamically stable under the temperature at which the outer carbonate would have been in equilibrium If we let V0 and V1 be the initial and residual volumes of with magnetite. In the absence of any plausible mechanism carbonate, V2 be the volume of magnetite, V3 be the void to enable a large temperature gradient between the two lay- space volume (since V > V + V ) and V ; V 0 1 2 FeCO3 MgCO3 ers, whose separation is only on the order of tens of mi- and V are the molar volumes of siderite, magnesite Fe3O4 crons, the sideritic composition of the residual carbonate and magnetite respectively (see Table 8) then from reaction in both layers should be very similar. The fact that they (20) it is easy to show that: are not, makes any subsequent inferences based on the data V 0 ¼ x0 V þð1 x0ÞV provided in the Treiman (2003) Table 1 questionable. Trei- FeCO3 MgCO3 man (2003) does suggest that the presence of other mineral V 1 ¼ a ðx1 V þð1 x1ÞV Þ FeCO3 MgCO3 phases in the carbonate grains analyzed could account for ð21Þ b such inconsistencies, however no other investigation of V 2 ¼ V 3 Fe3O4 ALH84001 carbonate has documented such phases at more V ¼ V V V than trace levels. This extreme difference in residual carbon- 3 0 1 2 ate composition can be used to completely negate the Trei- man (2003) equilibrium model for siderite decomposition. Table 8 4.2.2.6. Inferred volume abundances of magnetite in magne- Molar volumes of siderite, magnesite, and magnetite. tite-bearing layers. Assuming all the magnetite present in Species Vh (105 m3) the inner and outer magnetite rich rims are the products of partial thermal decomposition, and that these magnetites FeCO3 2.938 MgCO3 2.803 are essentially stoichiometric pure Fe3O4, then the initial composition of the carbonate prior to decomposition would Fe3O4 4.452 Author's personal copy

6670 K.L. Thomas-Keprta et al. / Geochimica et Cosmochimica Acta 73 (2009) 6631–6677

From the Eq. (21) the fractional volume of magnetite, magnetites. If such coating were the source of the observed V Fraction, in the ﬁnal residual carbonate–magnetite assem- Cr and/or Al then they would need to be relatively thick Fe3O4 blage is given by: and would be obvious, even in casual observation.

V V ðx0 x1Þ V Fraction ¼ 2 ¼ Fe3O4 ð22Þ Fe3O4 V 1 þ V 2 V ðx0 x1Þþ3 ð1 x0Þðx1 V Fe O þð1 x1ÞV Þ Fe3O4 3 3 MgCO3

Taking the inferred compositions of the initial and residual 4.4. Thermal decomposition studies of sideritic carbonates ALH84001 carbonate in the inner and outer rims from Ta- bles 1 and 2 of Treiman (2003) we find that the fractional Carbonate solid solutions between siderite and magne- volume of magnetite inferred to be present in the outer site, rhodochrosite, and calcite can be viewed as the inter- rim to be only 15% and only 43% in the inner rim (Table penetration, on an atomic scale, of two chemically 7). These values are inconsistent with any magnetite rim different but structurally similar lattices (i.e., calcite; R3c). abundances reported in the literature, based on our obser- For ionic systems such as the carbonate minerals, the limits vations, the volume fraction magnetite in these rims typi- of solubility generally correlate with the relative size of the cally lies in the range of 50–70%. This again suggests the ions and their electrical charge in accordance with the data provided by Tables 1 and 2 of Treiman (2003) are Hume–Rothery rules (Hume-Rothery et al., 1969). Since too inaccurate to be useful in elucidating the origin of mag- the ionic radii for octahedrally coordinated Mg2+,Mn2+, netite. This simple mass balance treatment shows that the and Fe2+ are all similar (Table 9) they demonstrate com- magnetite in the rims could not have formed from an initial plete solid solution series in essentially all geological envi- carbonate composition now reflected by the assumed post- ronments. In the case of Ca2+, where the mismatch in decomposition residual carbonate. The decomposition radii is greater than 10%, solid solutions within the calcite hypothesis cannot explain the data in Treiman (2003). crystal structure are only possible over a relatively limited compositional range. 4.3. Conflicting experimental observations 4.4.1. Prior studies Two sets of direct experimental observations indicate a Thermal decomposition of mixed cation (Mg,Ca,Mn)- significant fraction of magnetites in ALH84001 carbonate siderites have been investigated under a wide variety of con- arguably could not have formed by either the Brearley ditions, not only because of their geochemical importance, (2003) or Treiman (2003) models. These observations are: but also because they provide a convenient synthetic route to the preparation of complex metal oxides via the so-called Magnetites embedded within the magnesite layer sepa- ‘solid-solution-precursor’ method (Vidyasagar et al., 1984; rating the inner and outer magnetite rich bands (e.g., Vidyasagar et al., 1985). The results of such studies are Fig. 3G). summarized in Table 6 and invariably demonstrate that Chemically impure magnetites with minor to trace thermal decomposition yields a mixed metal oxide phase amounts of Cr and/or Al (Fig. 5C). as the product17. In general, substitution of Fe2+ by Mg2+ or Mn2+ in siderite increases its thermal stability In the first case, since the carbonate in the magnesite (Dubrawski, 1991a; Dubrawski, 1991b), generally to a far layer is essentially free of Fe, there is no siderite with which greater extent than thermodynamic factors alone could ac- to decompose to form magnetite. In the second case since count for. Furthermore, thermogravimetric analysis dem- neither Cr nor Al can substitute into the trigonal ðR3cÞ onstrates that decomposition is accompanied by only a structure of carbonate, thermal decomposition of carbonate single endothermic peak (Dubrawski, 1991b) indicating alone cannot form magnetites with these elemental impuri- the siderite component of the solid solution does not ties. While we note that Treiman (2003) does allude to the decompose independently. presence of “smectite or Fe-oxyhydroxide adhering to the grains’ irregularities”, we have not observed any evidence 4.4.2. Thermal decomposition of Roxbury siderite of such surface coating in any TEM imaging of embedded Roxbury siderite represents a good terrestrial analog to the siderite-rich carbonate compositions suggested by Trei- man (2003) to have decomposed to form the magnetite rich Table 9 bands of the ALH84001 carbonates. Lane and Christensen Crystal and ionic radii for octahedral Mg2+,Mn2+ and Fe2+. Cation (VIII coord.) Crystal radius (nm) Ionic radius (nm) 17 In this respect, it is worth re-emphasizing that not only do Mg2+ 0.103 0.089 (Mg,Mn)-siderites form a continuous substitutional solid solution Ca2+ 0.126 0.112 series (magnesiosiderite manganosiderite), but so do the interme- Mn2+ 0.111 0.096 diate (Mg,Mn)-Fe-monoxides (magnesiowustite manganowustite) Fe2+ 0.106 0.092 and product (Mg,Mn)-spinels (magnesioferrite manganoferrite). Author's personal copy

Origins of Magnetite in ALH84001 6671

(1997) reported the bulk composition of Roxbury siderite reported here19; both of which find that the thermal decom- as (Fe0.84Mg0.10Mn0.04Ca0.02)CO3, based on electron micro- position of a mixed cation siderite invariably results in the probe analyses at a maximum spatial resolution of 1 lm. formation of mixed cation magnetite. We suggest, however, High resolution TEM characterization at the sub-micron that the results obtained by Golden et al. (2006) are in fact scale however, indicates significant deviations in Fe:Mg ra- entirely consistent with prior literature and the Roxbury tio from the bulk composition, correlating with grain size. siderite studies – it is only the interpretation of these results The fine grain (<100 nm) component is Mg-poor by Golden et al. (2006) that is flawed. (Fe:Mg > 20:1) while, by comparison, the coarse grained To illustrate this, it is first necessary to review the mate- component is Mg-rich (Fe:Mg < 8:1) (see Fig. 6A). In both rials and methods used by Golden et al. (2006). The natural cases Mn and Ca appeared relatively uniform in siderite used for their experiments was from Copper Lake, composition. Antigonish County, Nova Scotia and is reported to have a Carbonate decomposition under both ‘slow’ and ‘fast’ bulk chemical composition of (Fe0.643Mg0.345Mn0.012)CO3 heating resulted in the formation of impure Fe-oxides or (Golden et al., 2006). This siderite was partially decom- ferrites (Figs. 6C and 6F). The only significant difference posed in an evacuated sealed quartz tube by controlled between the samples was the degree to which the carbonate heating, both in the presence of, and absence of, a natural had undergone decomposition. In the ‘slow’ heated sample pyrite containing minor silicate inclusions. The heating pro- no residual carbonate remained (Figs. 6B and 6C). In the file used was a slow temperature ramp (1 K min1) from ‘fast’ heated18 sample only the very upper surface of the sid- ambient up to 623 K for nine days before then cooling back erite exposed to laser heating underwent complete decom- again to ambient. Since the siderite only underwent partial position to ferrites (Figs. 6E and 6F), while the decomposition to magnetite the residual carbonate was re- underlying carbonate phase, although heavily vesiculated, moved by dissolution in “20% acetic acid followed by mul- contained no ferrites (Fig. 6G). This observation is in sharp tiple washings with distilled H2O.” The chemical contrast to the magnetite rich rims and veins of ALH84001 composition of the remaining magnetite was then deter- carbonate disks that are characterized by an intimate mix- mined using TEM/EDX. In their results they report that ture of magnetites embedded in a host carbonate matrix. Copper Lake siderite in the absence of pyrite produced The absence of any co-mixed carbonate/ferrite zone in magnetite containing only 7.1 mol.% Mg, while siderite in our experiments is consistent with the proposed decomposi- the presence of pyrite produced chemically pure magnetite tion mechanism of siderite involving the endothermic for- (<1 mol.% Mg). At this point it is necessary to inject a mation of a wu¨stite (FeO) intermediate that is rapidly few observations. It is clear that while the bulk composition oxidized in an exothermic reaction to produce the Fe3O4 of Copper Lake siderite may contain 34.5 mol.% Mg it can- product. not have been homogeneously distributed, since if it was In products produced by both ‘slow’ and ‘fast’ heating, there would be a single decomposition temperature associ- the Mg content of the mixed ferrite varied from one crystal ated with the siderite, and after nine days it would have to another (Figs. 6C and 6F) while the Mn content either all decomposed or not decomposed at all. Though remained relatively constant. However, since this composi- it could be argued that re-equilibration, by solid state cat- tional distribution appeared to mirror that of the ion diffusion, of Mg between magnetite product and unre- unheated carbonate we believe this is simply a reflection acted carbonate could have lead to the thermal of the initial Mg and Mn content of the carbonate (notably, stabilization of the residual carbonate by increasing its in the product phases produced by both ‘slow’ and ‘fast’ Mg mole fraction, such cation diffusion at 623 K would heating we did not find any evidence for discrete periclase not have been sufficiently facile to allow such a process or lime). (as discussed earlier). Fortunately, the issue of the compositional homogeneity of Copper Lake siderite was reported 4.4.2.1. Comparison to decomposition studies of Copper Lake previously by Bell (2007), who is affiliated with the same re- siderite. Golden et al. (2006) have recently reported they search group as Golden et al. (2006) and so presumably were able to produce almost pure nanophase magnetite used the same parent siderite sample.20 Fig. 7F taken from from the slow thermal decomposition of a natural siderite Fig. 3A, 3B–3G of Bell (2007) illustrates the variation of from Copper Lake, Nova Scotia. From these results, they Mg and Fe abundances in a single grain surface have argued for an “exclusively inorganic origin” for mag- 175 175 lm of Copper Lake siderite. In this single re- netite in ALH84001. Clearly these results would appear to gion, there are at least three distinct carbonate composi- be in stark conflict with both prior literature (see Table tions as illustrated in the histogram shown in Fig. 7F. 6), and the results from the Roxbury siderite studies More importantly, one of these carbonate compositions is nearly pure siderite with only minor to trace Mg. Turning back to the decomposition experiments of Golden et al. 18 “The use of high-power lasers with short wavelengths allows (2006) it is clear that decomposition of Copper Lake generation of the most extreme shock pressures in laboratories, but only in conjunction with extremely short pulse durations in the nanoseconds range. Laser irradiation shock experiments may be 19 The “slow” heating studies of Roxbury siderite were conducted regarded as adequate simulations of hypervelocity impacts of under essentially analogous conditions to those used by Golden microgram-mass micrometeorites onto atmosphere-free bodies et al. (2006). such as the Moon as well as onto spacecraft” from Langenhorst 20 This is consistent with the observation that both report identical et al. (2002). compositions for their bulk Copper Lake siderite. Author's personal copy

6672 K.L. Thomas-Keprta et al. / Geochimica et Cosmochimica Acta 73 (2009) 6631–6677

Fig. 7F. Compositional variability of siderite–magnesite solid solutions in a single grain of Copper Lake siderite, data taken from Figs. 3b and c of Bell (2007). The image on the left is a Hue–Saturation–Intensity (HSI) element ratio map for Fe:Mg (see Bensen and Lechene (2005)). The respective Fe and Mg signal intensities are derived from the grey-scale image pixel values of the electron microprobe Fe and Mg element maps shown in Fig. 3b and c of Bell (2007). Since neither electron microprobe map has an accompanying calibrated intensity scale the Fe:Mg values shown for the HSI image represents relative values and so are not absolute. In the HSI image, each pixel has a color determined by value of the Fe:Mg ratio, while the intensity, or brightness, is proportional to the signiﬁcance of that ratio as determined by the combined intensities of the Fe and Mg pixels in the respective electron microprobe maps. Pixels with low intensities appear darker and pixels with high intensities appear brighter. Three distinct solid-solution compositions for carbonate are apparent. This includes an almost pure siderite component colored pink-purple which is encapsulated by two, more magnesite rich, solid solutions colored blue and green. On the right is a normalized histogram of the Fe:(Fe + Mg) ratios calculated for each grey-scale pixels in the respective electron microprobe maps. Again three distinct distributions are clearly apparent as indicated by the three Gaussian curves shown in green. As with the HSI image the values of Fe:(Fe + Mg) ratios are relative and not absolute. The bulk composition of Copper Lake siderite has been previously reported as

(Fe0.643Mg0.345Mn0.012)CO3 (Golden et al., 2006). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this paper.) siderite will thus occur in a number of stages with Keprta et al., 2000a). Clearly these chemically impure increasing temperature, corresponding to the successive magnetites have detrital origins and did not form as a decomposition of progressively more Mg-rich carbonate consequence of thermal decomposition of host carbonate components. At a temperature of 623 K only the stability which lacks detectable Cr and/or Al. field of the most Mg-poor siderite would have been crossed, Although the “thermal decomposition” hypothesis for resulting in the formation of a magnetite product that pre- the observation of chemically pure magnetites in the rims dictably has only a low Mg abundance. This is completely of ALH84001 carbonate disks appears, on the surface, to consistent with prior literature and the siderite decomposi- provide a simple inorganic explanation of the observations, tion studies reported here. That Golden et al. (2006) failed theoretical and experimental evaulation indicates it is not to fully characterize, or even take into consideration, the applicable to the formation of the vast majority of the composition of their starting materials or determine the magnetites in ALH84001 carbonate disks (Table 10). As fraction of siderite that underwent decomposition or the discussed at length previously, both the Brearley (2003) fraction and composition of the residual carbonate, seem and Treiman (2003) hypotheses proposed to explain the to invalidate any of their interpretations or conclusions. presence of ALH84001 magnetite are flawed. The importance of experimental data therefore cannot be overesti- 4.5. Implications and origins of ALH84001 magnetites mated. We can summarize the experimental studies of decomposition of a mixed cation siderite with the following The majority of ALH84001 magnetites are stoichiomet- generalized reaction: rically pure Fe3O4 embedded within all compositions of 3ðFe Mg Ca Mn ÞCO carbonate ranging from the most Fe-rich ((Fe0.60Mg0.16 x y z ½1xyz 3 D Ca0.05Mn0.04)CO3) at the boundary of the inner and outer !ðFexMgy CazMn½1xyzÞ3O4 þ 2CO2 þ CO ð23Þ cores to the most Fe-free magnesite ((Mg0.95Ca0.05)CO3). It is difficult to suggest a process by which this magnesite While it is impossible to completely discount that some of has undergone thermal decomposition to produce chemi- the magnetites embedded in ALH84001 carbonates might cally pure magnetite based on results from experiments in be a product of thermal decomposition, we argue at best Table 6 and those of the Roxbury siderite decomposition this can only account for a tiny portion. Consequently, this studies reported here. Only a small fraction of Martian should not divert our attention from understanding the ori- magnetite crystals are chemically impure; they contain trace gin of the remaining majority of magnetites that would ap- to minor amounts of Cr (Fig. 5C) and/or Al (Thomas- pear to have been derived from a detrital or allochthonous Author's personal copy

Origins of Magnetite in ALH84001 6673

Table 10 Features in ALH84001 inconsistent with carbonate decomposition. ALH84001 observations Incompatibility with partial thermal carbonate decomposition Micron scale isotopic and chemical heterogeneity present in Increased cation diffusion at high temperatures would have erased carbonate disks, e.g., Valley et al. (1997), Saxton et al. (1998), Eiler chemical and isotopic gradients observed in the carbonate. et al. (2002), McKay et al. (1996) and this study 40Ar/39Ar chronometer in ALH84001 feldspathic glass last reset Heating above the Ar release temperature, 350–500 °C, after 4 Ga 4Ga(Weiss et al., 2002) would have reset 40Ar/39Ar chronometer. Chemically pure magnetites present within all ALH84001 carbonate Breakdown temperature for thermal decomposition of Fe-containing compositions including those contain no Fe (this study) carbonates to magnetite is a function of cation composition; the differing carbonate compositions in ALH84001 require different and incompatible temperature constraints. Magnetite cation composition reflects that of the carbonate from which it forms, i.e., chemically pure magnetite requires decomposition of chemically pure siderite. Magnetites present within the magnesite rim cannot have formed by magnesite decomposition since it contains no Fe. A fraction of magnetites present in rim and core carbonate contain Cr and Al cations cannot substitute in to Fe–Mg–Ca carbonate, minor Cr and/or Al, e.g., Thomas-Keprta et al. (2000a) and this therefore the Cr and/or Al containing magnetites could not have study. formed from decomposition of host carbonate. Absence of graphite within carbonate disks, e.g., McKay et al. (1996), Under thermodynamic equilibrium graphite is a preferred product of Flynn et al. (1998), Flynn et al. (1999), Clemett et al. (1998) and this siderite decomposition. Presence of nanophase magnetite catalyst study. would help to lower kinetic barriers for graphite precipitation. Some magnetites embedded within carbonate show evidence of Volume change associated with carbonate decomposition to surrounding void space, e.g., Barber and Scott (2002), Brearley (2003) magnetite too large to account for observed void sizes, while many and this study. magnetites lack any association with void spaces. source (Table 11). There are no convincing data showing 5. SUMMARY AND CONCLUSIONS that they originated from the carbonate. Although we have not attempted to present a detailed model for a detrital ori- Two heating scenarios described herein provide the geo- gin of ALH84001 magnetite, we have highlighted the pres- logical context for the decomposition models developed ence of many features consistent with exposure of the independently by Brearley (2003) and Treiman (2003). carbonates to aqueous fluids after formation (Table 11). These models are contradictory; that is, they cannot both

Table 11 ALH84001 features in ALH84001 consistent with allochthonous magnetite. ALH84001 observations Implications Chemically pure magnetites embedded within a diverse range of Embedding carbonate is not the progenitor of the magnetite – detrital carbonate compositions, including magnesite that contains no Fe magnetite incorporated during disk formation. (this study) Fine-grained, S-containing, veins that cross-cut core carbonate (this Influx of a sulfurous fluid(s) after carbonate disk formation. study). Ferrihydrite embedded within fissures in the magnesite band (this Influx of ferrous–ferric fluid(s) after carbonate disk formation. study). Magnetites containing minor Cr and/or Al present in rim and core Embedding carbonate is not the progenitor of the magnetite – detrital carbonate (Thomas-Keprta et al. (2000a) and this study). magnetite incorporated during disk formation. Amorphous Si present in nanophase carbonate magnetite-rich rim Influx of siliceous fluid(s) either concurrent with or subsequent to and veins (this study). rim/vein formation. Spatially association of abundant Fe-sulfides with magnetite-rich Influx of a sulfurous fluid(s) after carbonate disk formation. rims but their relative scarcity within magnetite-rich veins (this study). Presence of void space associated with some magnetites embedded Carbonate recrystallization enables migration of imperfections, such within both core carbonate and the magnesite band (Barber and Scott as magnetite and void spaces are to grain boundaries. (Conversely (2002), Brearley (2003), and this study). Presence of void space such spatial associations may simply be coincidental, e.g., void spaces associated with some magnetites located at the interface of carbonate are associated with magnetites in the magnesite band, but these grains (this study). magnetites could not have been produced by magnesite thermal decomposition.) Author's personal copy

6674 K.L. Thomas-Keprta et al. / Geochimica et Cosmochimica Acta 73 (2009) 6631–6677 have occurred since application of one model negates the International Conference on Secondary Ion Mass Spectrometry, applicability of the other. The first is based on carbonate Manchester, UK, p. 269. decomposition occurring under “extreme disequilibrium Borg L. E., Connelly J. N., Nyquist L. E., Shih C.-Y., Wisemann conditions” in which “kinetics are the dominant controlling H. and Reese Y. (1999) The age of carbonates in Martian factor” determining the chemical and physical nature of the meteorite ALH84001. Science 286, 90–94. Bradley J. P., Harvey R. P. and McSween J. H. Y. (1996) magnetites that are formed. Although never explicitly ad- Magnetite whiskers and platelets in the ALH84001 Martian dressed, this model is most consistent with the impact event meteorite: evidence of vapor phase growth. Geochim. Cosmo- that ejected ALH84001 from Mars. In the second model, chim. Acta 60, 5149–5155. carbonate decomposition occurs “at some depth beneath Brearley, A. J. (1998) Rare potassium-bearing mica in Allan Hills the Martian surface where the pressure was greater than 84001: additional constraints on carbonate formation. Martian the atmospheric pressure and the temperature declined meteorites: where do we stand and where are we going? Abstracts slowly.” However, based on kinetic and thermodynamic from a workshop,p.6. arguments, both models proposed for the high temperature, Brearley A. J. (2003) Magnetite in ALH 84001: an origin by shock- inorganic formation of ALH84001 magnetite would not induced thermal decomposition of iron carbonate. Meteorit. have produced the results observed in ALH84001 carbon- Planet. Sci. 38, 849–870. Cady S. L., Wenk H. R. and Downing K. H. (1996) HRTEM of ate disks. microcrystalline opal in chert and porcelanite form Monterey ALH84001 carbonate assemblages can be best explained Formation, California. Am. Mineral. 81, 1380–1395. as the result of low temperature, disequilibrium precipita- Chai L. and Navrotsky A. (1996a) Synthesis, characterization and tion from a single fluid with variable composition or from enthalpy of mixing of the (Fe,Mg)CO3 solid solution. Geochim. multiple fluids. Nanophase magnetite and Fe-sulfides were Cosmochim. Acta 60, 4377–4383. suspended in fluids that formed the disk cores and rims. Chai L. and Navrotsky A. (1996b) Synthesis, characterization, and After deposition ALH84001 carbonate disks were exposed energetics of solid solution along the dolomite-ankerite join, to multiple fluids containing amorphous silica, additional and implications for the stability of ordered CaFe(CO3)2. Am. nanophase magnetite, and S- and Fe-rich phases, some of Mineral. 81, 1141–1147. which were deposited in veins. While the majority of Clemett, S. J., Dulay, M. T., Gillette, J. S., Chillier, X. D. F., Mahajan, T. B. and Zare, R. N. (1998) Evidence for the ALH84001 magnetites were deposited in silica enriched extraterrestrial origin of polycyclic aromatic hydrocarbons in rims and veins, some are also distributed throughout the the Martian meteorite ALH84001. Chemistry and Physics of cores and within the magnesite bands. Most magnetites Molecules and Grains in Space. faraday Discussions No. 109, pp. are chemically pure, although a few contain minor Al/Cr. 417–436. Their presence is inconsistent with formation by thermal Cohn, A. (2006) Formation of magnetite nanoparticles by thermal decomposition of their host carbonate. We suggest that decomposition of iron bearing carbonates: implications for the the majority of ALH84001 magnetites has an allochtho- evidence of fossil life on Mars. In National Nanotechnology nous origin and was added to the carbonate system from Infrastructure Network, Reseach Experience for Undergraduates, an outside source. This origin does not exclude the possibil- Accomplishments Program, Cornell, NY, USA, pp. 58–59. ity that a fraction is consistent with formation by biogenic de La Croix A., English R. B. and Brown M. E. (1998) Modeling the thermal decomposition of solids on the basis of lattice processes, as proposed in previous studies. energy changes. J. Solid State Chem. 137, 332–345. Dhupe A. P. and Gokarn A. N. (1990) Studies in the thermal ACKNOWLEDGEMENTS decomposition of natural siderites in the presence of air. Int. J. Miner. Process. 20, 209–220. We acknowledge support from the NASA Mars Fundamental Domeneghetti M. C., Fioretti A. M., Ca´mara F., Molin G. and program. We thank FEI Corporation, particularly Trisha Rice Tazzoli V. (2007) Thermal history of ALH 84001 meteorite by and Lucille Giannuzzi, for assistance in preparing the FIB sections Fe2+-Mg ordering in orthopyroxene. Meteorit. Planet. Sci. 42, of ALH84001 carbonate disks. We also acknowledge Giles Graham 1703–1710. for assistance with preparing the Texas FIB sections. We gratefully Dubrawski, J. V. (1991a). Differential scanning calorimetry and its acknowledge the support of Sean and Nathaniel Keprta and Miss B. applications to mineralogy and the geosciences. In: Thermal and O. Clemett. We thank two anonymous reviewers and Christian Analysis in the Geosciences, Lecture Notes in Earth Sciences Koeberl for suggestions that greatly improved this paper. (eds. W. Smykatz-Kloss and S. St. J. Warne). Springer, Berlin/ Heidelberg, pp. 16–59. Dubrawski J. V. (1991b) Thermal decomposition of some siderite- REFERENCES magnesite minerals using DSC. J. Therm. Anal. 37, 1213–1221. Eiler J. M., Valley J. W., Graham C. M. and Fournelle J. (2002) Bagin V. I., Gendler T. S., Rybak R. S. and Kut’min R. N. (1974) Two populations of carbonate in ALH84001: geochemical Transformation temperatures of the natural solid solution evidence for discrimination and genesis. Geochim. Cosmochim. ðFe; MgÞCO3. Izv. Russ. Acad. Sci. Phys. Solid Earth 6, 73–84. Acta 66, 1285–1303. Barber D. J. and Scott E. R. D. (2002) Origin of supposedly Fisler D. K. and Cygan R. T. (1998) Cation dissolution in calcite: biogenic magnetite in the Martian meteorite Allan Hills 84001. determining closure temperatures and the thermal history for the Proc. Natl. Acad. Sci. USA 99, 6556–6561. Allan Hills 84001 meteorite. Meteorit. Planet. Sci. 33, 785–789. Bell M. S. (2007) Experimental shock decomposition of siderite and Flynn G. J., Keller L. P., Jacobsen C. and Wirick S. (1998) Carbon the origin of magnetite in Martian meteorite ALH84001. in Allan Hills 84001 carbonate and rim. Meteorit. Planet. Sci. Meteorit. Planet. Sci. 42, 935–949. 33, A50–A51. Bensen, D. and Lechene, C. P. (2005) Quantitating MIMS ratios Flynn, G. J., Keller, L. P., Jacobsen, C. and Wirick, S. (1999) using Hue, Saturation, and Intensity displays. In Fifteenth Organic carbon in Mars meteorites: a comparison of Author's personal copy

Origins of Magnetite in ALH84001 6675

ALH84001 and Nakhla. Lunar Planet. Sci. XXX Lunar Planet. Hume-Rothery W., Smallman R. W. and Haworth C. W. (1969) Inst., Houston. #1087 (abstr.). The Structure of Metals and Alloys. Metals & Metallurgy Trust, Folk, R. L. (1965) Some aspects of recrystallization in ancient London. limestones. Society of Economic Paleontologists and Mineralo- Hurst H. J., Levy J. H. and Patterson J. H. (1993) Siderite gists Special publication 13, pp. 14–48. decomposition in retorting atmospheres. Fuel 72, 885–890. Forster B., Van de Ville B., Berent J., Sage D. and Unser M. (2004) Isambert A., De Resseguier T., Gloter A., Reynard B., Guyot F. Complex wavelets for extended depth-of-field: a new method and Valet J.-P. (2006) Magnetite-like nanocrystals fromed by for the fusion of multichannel microscopy images. Microsc. laser-driven shocks in siderite. Earth Planet. Sci. Lett. 243, 820– Res. Techniq. 65, 33–42. 827. French B. M. and Eugster H. P. (1965) Experimental control of Isambert A. and Valet J.-P. (2003) Stable Mn-magnetite derived oxygen fugacities by graphite-gas equilibriums. J. Geophys. Res. from Mn-siderite by heating in air. Earth Planet. Sci. Lett. 108, 70, 1529–1539. EMP2-1–EMP2-9. Fritz J., Artemieva N. and Greshake A. (2005) Ejection of Martian Iwafuchi K., Watanabe C. and Otsuka R. T. (1983) Thermal meteorites. Meteorit. Planet. Sci. 40, 1393. decomposition of magnesian kutnahorite. Thermochim. Acta Fritz J., Greshake A. and Soffler D. (2004) Micro-raman spectros- 60, 361–381. copy of plagioclase and maskelyinte in Martian meteorites: Jagtap S. B., Pande A. R. and Gokarn A. N. (1992) Kinetics of evidence of progressive shock metamorphism. Antarct. Mete- thermal decomposition of siderite: effect of particle size. Int. J. orit. Res. 18, 96–116. Miner. Process. 36, 113–124. Fritz, J., Greshake, A. and Sto¨ffler, D. (2003) Launch conditions Jurewicz, A. J. G., Lofgren, G. E., Williams, R. J., Lanier, A., Le, for Martian meteorites: plagioclase as a shock pressure L., Carter, W., Wagstaff, J. and Roshko, A. (1993) NASA barometer. Lunar Planet. Sci. XXXIV. Lunar Planet. Inst., Johnson Space Center Technical Update: JSC system unsing a Houston. #1355 (abstr.). solid electrolytic cell in a remote location to measure oxygen

Gallagher P. K. and Warne S. S. J. (1981a) Application of fugacities in CO/CO2 controlled-atmosphere furnaces, Hous- thermomagnetometry to the study of siderite. Mater. Res. Bull. ton, TX, USA. 16, 141–144. Karl S. and Deborah G. (2004) Petrographic and geochemical Gallagher P. K. and Warne S. S. J. (1981b) Thermomagnetometry analysis of detrital magnetite of last Wisconsinan tills in eastern and thermal decomposition of siderite. Thermochim. Acta 43, Indiana and western Ohio. Geolog. Soc. Am. Abstr. Prog. 36, 253–267. 48. Galwey A. K. and Brown M. E. (1999) Thermal Decomposition of Kennedy L. A. and White J. C. (2001) Low-temperature recystal- Ionic Solids, 86: Studies in Physical and Theoretical Chemistry. lization in calcite: mechanisms and consequences. Geology 29, Elsevier Science, Amsterdam. 1027–1030. Gasem K. A. M., Gao W., Pan Z. and Robinson R. L. J. Kent A. J. R., Hutcheon I. D., Ryerson F. J. and Phinney D. L. (2001) A modified temperature dependence for the Peng- (2001) The temperature of formation of carbonate in Martian Robinson equation of state. Fluid Phase Equilib. 181, 113– meteorite ALH84001: constraints from cation diffusion. Geo- 125. chim. Cosmochim. Acta 65, 311–321. Geider S., Dussol B., Nitsche S., Veesler S., Berthe´ze`ne P., Dupuy Kirschvink J. L., Maine A. T. and Vali H. (1997) Paleomagnetic P., Astier J. P., Boistelle R., Berland Y., Dagorn J. C. and evidence of a low-temperature origin of carbonate in the Verdier J. M. (1996) Calcium carbonate crystals promote Martian meteorite ALH84001. Science 275, 1629–1633. calcium oxalate crystallization by heterogeneous or epitaxial Koziol, A. M. (2001) A siderite–magnesite decarbonation study. nucleation: possible involvement in the control of urinary Geological Society of America National Meeting, Boston. lithogenesis. Calcified Tissue Int. 59, 33–37. #25840 (abstr.). Giannuzzi L. A. and Stevie F. A. (1999) A review of focused ion Koziol, A. M. and Brearley, A. J. (2002) A non-biological origin beam milling techniques for TEM specimen preparation. for the nanophase magnetite grains in ALH84001: experimental Micron 30, 197–204. results. Lunar Planet. Sci. XXXIII. Lunar Planet. Inst., Hous- Golden, D. C., Ming, D. W., Lauer, H. V. J., Morris, R. V., ton. #1672 (abstr.). Treiman, A. H. and McKay, G. A. (2006) Formation of Kubas, Z. and Szalkowicz, M. (1971) Application of simultaneous “chemically pure magnetite from Mg–Fe carbonates: impli- differential thermal analysis, thermogravimetry and analysis of cations for the exclusively inorganic origin of magnetite evolved gases to studies on kinetics and mechanism of complex and sulfides in Martian meteorite ALH84001. Lunar reactions. Thermal Analysis (Proc. of the 3rd ICTA, Davos, Planet. Sci. XXXVII. Lunar Planet. Inst., Houston. 1971) 2, 447–461. #1199 (abstr.). Lane M. D. and Christensen P. R. (1997) Thermal infrared Gotor F. J., Macias M., Ortega A. and Criado J. M. (2000) emission spectroscopy of anhydrous carbonates. J. Geophys. Comparative study of the kinetics of the thermal decomposition Res. 102, 25581–25592. of synthetic and natural siderite samples. Phys. Chem. Miner. Langenhorst F., Boustie M., Deutsch A., Hornemann U., Mati- 27, 495–503. gnon C., Migault A. and Romain J. P. (2002) Experimental Hakonardottir, K. M., Hogg, A. J. and Johannesson, T. (2002) techniques for the simulation of shock metamorphism: a case Flow of a granular avalanche over an obstacle. Eur. Geophys. study for calcite. In High-Pressure Shock Compression of Solids Union Gen. Assembly XXVII. #1925 (abstr.). V: Shock Chemistry with Application to Meteorite Impacts (eds. Harvey R. P. and McSween J. H. Y. (1996) A possible high- L. Davison, Y. Horie and T. Sekine), pp. 1–27. High-Pressure temperature origin for the carbonates in the Martian meteorite Shock Compression of Solids V: Shock Chemistry with ALH84001. Nature 382, 49–51. Application to Meteorite Impacts. Springer Materials, Berlin. Holland T. and Powell R. (1998) An internally-consistent thermo- Langenhorst, F., Shaw, C. S. J. and Metzler, K. (2000) Mineral dynamic dataset for phases of petrological interest. J. Meta- chemistry and microstructures in ALH84001. Lunar Planet. Sci. morph. Geol. 16, 309–343. XXXI. Lunar Planet. Inst., Houston. #1866 (abstr.). Huebner J. S. (1975) Oxygen fugacity values of furnace gas Lauer, H. V., Ming, D. W. and Golden, D. C. (2005) Thermal

mixtures. Am. Mineral. 60, 815–823. characterization of Fe3O4 nanoparticles formed from poorly Author's personal copy

6676 K.L. Thomas-Keprta et al. / Geochimica et Cosmochimica Acta 73 (2009) 6631–6677

crystalline siderite. Lunar Planet. Sci. XXXVI. Lunar Planet. northwestern Japan sea for provenance study. J. Sediment. Res. Inst., Houston. #2153 (abstr.). 62, 802–809. Lewis G. N. and Randall M. (1961) Thermodynamics, seconnd ed. Reid R. P. and Macintyre I. G. (1998) Carbonate recyrstallization McGraw-Hill, New York. in shallow marine environments: a widespread diagenitic Mathews, J. H. and Fink, K. D. (2004) Solution of diﬀerential process forming micritized grains. J. Sediment. Res. 68, 928– equations: predictor–corrector methods. In: Numerical Methods 946. using Matlab (fourth ed.) Prentice-Hall Inc., Upper Saddle Romanek C. S., Grady M. M., Wright I. P., Mittlefehldt D. W., River, pp. 505–508. ISBN 0-13-065248-2. Socki R. A., Pillinger C. T. and Gibson J. E. K. (1994) Record McCauley J. W. and Roy R. (1974) Controlled nucleation and of ﬂuid-rock interactions on Mars from the meteorite

crystal growth of various CaCO3 phases by the silica gel ALH84001. Nature 372, 655–657. technique. Am. Mineral. 59, 947–963. Saxton J. M., Lyon I. C. and Turner G. (1998) Correlated chemical McKay D. S., Gibson J. E. K., Thomas-Keprta K. L., Vali H., and isotopic zoning in carbonates in the Martian meteorite Romanek C. S., Clemett S. J., Chiller X. D. F., Maechling C. R. ALH84001. Earth Planet. Sci. Lett. 160, 811–822. and Zare R. N. (1996) Search for past life on Mars: possible Scott E. R. D., Krot A. N. and Yamaguchi A. (1997) Formation of relic biogenic activity in Martian meteorite ALH84001. Science carbonates in Martian meteorite ALH84001 from shock melts. 273, 924–930. Meteorit. Planet. Sci. 32, A117–A118. McKay G., Schwandt C. and Mikouchi T. (1998) Feldspathic glass Shuster D. L. and Weiss B. P. (2005) Martian surface paleotem- and silica in Allan Hills 84001. Meteorit. Planet. Sci. 33, A102. peratures from thermochronology of meteorites. Science 309, McSween, J. H. Y. (2002) Basalt or andesite? A critical evaluation 594–597. of constraints on the composition of the ancient Martian crust. Sto¨ffler D. (1982) Density of minerals and rocks under shock Lunar Planet. Sci. XXXIII. Lunar Planet. Inst., Houston. compression. In Landolt-Bo¨rstein: Numerical Data and Func- #1062 (abstr.). tional Relationships in Science and Technology (ed. K.-H. McSween J. H. Y. and Harvey R. P. (1998) An evaporation model Hellwege), pp. 120–183. Landolt-Bo¨rstein: Numerical Data for formation of carbonates in the ALH84001 Martian mete- and Functional Relationships in Science and Technology. orite. Intl. Geol. Rev. 40, 774–782. Springer, Berlin/Heidelberg/New York. McSween J. H. Y. and Treiman A. H. (1998) Martian meteorites. Sto¨ffler, D. (2000) Maskelynite confirmed as diaplectic glass: In Planetary Materials: Reviews in Mineralogy (ed. J. J. Papike). indication for peak shock pressures below 45 GPa in all Mineralogical Sociecty of America, Chantilly (chapter 6). Martian meteorites. Lunar Planet. Sci. XXXI. Lunar Planet. McSwiggen P. L. (1993a) Alternative solution model for the Inst., Houston. #1170 (abstr.).

ternary carbonate system CaCO3–MgCO3–FeCO3. Part 1. A Tamaura Y. and Tabata M. (1990) Complete reduction of carbon ternary Bragg–Williams ordering model. Phys. Chem. Miner. dioxide to carbon using cation-excess magnetite. Nature 346, 20, 33–41. 255–256. McSwiggen P. L. (1993b) Alternative solution model for the Thomas-Keprta K. L., Bazylinski B. A., Kirchvink J. L., Clemett S.

ternary carbonate system CaCO3–MgCO3–FeCO3. Part 2. J., McKay D. S., Wentworth S. J., Vali H., Gibson J. E. K. and Calibration of a combined ordering model and mixing model. Romanek C. S. (2000a) Elongated prismatic magnetite crystals Phys. Chem. Miner. 20, 42–55. in ALH84001 carbonate globules: potential Martian magneto- Mikouchi, T., Miyamoto, M., Koizumi, E., Makishima, J. and fossils. Geochim. Cosmochim. Acta 64, 4049–4081. McKay, G. A. (2006) Relative burial depths of Nakhalites: an Thomas-Keprta K. L., Clemett S. J., Bazylinski D. A., update. Lunar Planet. Sci. XXVII. Lunar Planet. Inst., Hous- Kirschvink J. L., McKay D. S., Wentworth S. J., Vali H., ton. #1865 (abstr.). Gibson J. E. K., McKay M. F. and Romanek C. S. (2001) Milodowski A. E., Morgan D. J. and Warne S. S. J. (1989) Thermal Truncated hexa-octahedral magnetite crystals in ALH84001: analysis of studies of the dolomite–ferroan dolomite–ankerite presumptive biosignatures. Proc. Natl. Acad. Sci. USA 98,

series. II. Decomposition mechanism in ﬂowing CO2 atmo- 2164–2169. sphere. Thermochim. Acta 152, 279–297. Thomas-Keprta K. L., Clemett S. J., Bazylinski D. A., Kirschvink Min K. and Reiners P. W. (2006) High-temperature Mars-to-Earth J. L., McKay D. S., Wentworth S. J., Vali H., Gibson J. E. K. transfer of meteorite ALH84001. Earth Planet. Sci. Lett. 260, and Romanek C. S. (2002) Magnetofossils from ancient Mars: a 72–85. robust biosignature in the Martian meteorite ALH84001. Appl. Mittlefehldt, D. W. (1994a) ALH84001 cumulate orthopyroxenite; Environ. Microbiol. 68, 3663–3672. a previously unappreciated Martian meteorite. Lunar Planet. Thomas-Keprta, K. L., Wentworth, S. J., McKay, D. S. and Sci. XXV. Lunar Planet. Inst., Houston. pp. 911–912. Gibson, J. E. K. (2000b) Field emission gun scanning electron Mittlefehldt D. W. (1994b) ALH84001, a cumulate orthopyroxe- microscopy (FEGSEM) and transmission electron (TEM) nite member of the Martian meteorite clan. Meteoritics 29, 214– microscopy of phyllosilicates in Martian meteorites 221. ALH84001, Nakhla, and Shergotty. Lunar Planet. Sci. XXXI. Muxworthy A. R. and Williams W. (2008) Critical superparamag- Lunar Planet. Inst., Houston. #1690 (abstr.). netic/single-domain grain sizes in interacting magnetite parti- Treiman, A. H. (1997) Chemical disequilibrium in carbonate cles: implications for magnetosome crystals. J. R. Soc. minerals of Martian meteorite ALH 84001: inconsistent with Interface, ISSN 1742-5689. high formation temperature. Lunar Planet. Sci. XXVIII. Lunar Nyquist L. E., Bogard D. D., Shih C.-Y., Greshake A., Stoﬄer D. Planet. Inst., Houston. #1192 (abstr.). and Eugster O. (2001) Ages and geologic histories of Martian Treiman A. H. (1998) The history of ALH 84001 revised: multiple meteorites. Space Sci. Rev. 96, 105–164. shock events. Meteorit. Planet. Sci. 33, 753–764. Powell R., Holland T. and Worley B. (1998) Calculating phase Treiman A. H. (2003) Submicron magnetite grains and carbon diagrams involving solid solutions via non-linear equations compounds in Martian meteorite ALH84001: inorganic, abiotic with examples using THERMOCALC. J. Metamorph. Geol. 16, formation by shock and thermal metamorphism. Astrobiology 577–588. 3, 369–392. Razjigaeva N. G. and Naumova V. V. (1992) Trace element Trunin R. F., Gudarenko L. F., Zhernokletov M. V. and Simakov composition of detrital magnetite from coastal sediments of G. V. (2001) Experimental Data on Shock Compression and Author's personal copy

Origins of Magnetite in ALH84001 6677

Adiabatic Expansion of Condensed Matter. Russian Federal and nitrate solid-solution precursors. J. Solid State Chem. 58, Nuclear Center-VNIIEF, Sarov, Russia. 29–37. Valley J. W., Eiler J. M., Graham C. M., Gibson, Jr., E. K., Romanek Viti C. and Frezzotti M.-L. (2001) Transmission electron microscopy C. S. and Stolper E. M. (1997) Low-temperature cabonate applied to ﬂuid inclusion investigations. Lithos 55, 125–138. concretions in the Martian meteorite ALH84001: evidence from Ware S. S. J. and French D. H. (1984) The application of stable isotopes and mineralogy. Science 275, 1633–1638. simultaneous DTA and TG to some aspects of oil shale van der Bogert, C. H., Schultz, P. H. and Spray, J. G. (1999) mineralogy. Thermochim. Acta 76, 179–200. Experimental frictional heating of dolomitic marble: new Warren P. H. (1998) Petrologic evidence for low-temperature, insights for Martian meteorite Allan Hills 84001. Lunar Planet. possibly ﬂood-evaporitic origin of carbonates in the ALH84001 Sci. XXX. Lunar Planet. Inst., Houston. #1970 (abstr.). meteorite. J. Geophys. Res. [Planets] 103, 16759. Velbel M. A. (2007) Surface textures and dissolution processes of Weiss B. P., Kirchvink J. L., Baudenbacher F. J., Vali H., Peters N. heavy minerals in the sedimentary cycle: examples from T., MacDonald F. A. and Wikswo J. P. (2000) A low pyroxenes and amphiboles. In Heavy Minerals in Use (eds. M. temperature transfer of ALH84001 from Mars to Earth. A. Mange and D. T. Wright). Elsevier, Doorwerth, The Science 290, 791–795. Netherlands, pp. 113–150. Weiss B. P., Shuster D. L. and Stewart S. T. (2002) Temperatures Velbel M. A., Wentworth S. J., Thomas-Keprta K. L., Donatelle on Mars from 40Ar/39Ar thermochronology of ALH84001. A. R. and McKay D. S. (2007) Microdenticles: aqueous Earth Planet. Sci. Lett. 201, 465–472. corrosion textures on weathered chain silicates as terrestrial Zakharov V. Y. and Adonyi Z. (1986) Thermal decomposition analogs of pyroxene alteration in Mars meteorites. Meteorit. kinetics of siderite. Thermochim. Acta 102, 101–107. Planet. Sci. 42, 5172. Zenobi R., Zhan Q. and Voumard P. (1995) Multiphoton Vidyasagar K., Gopalakrishnan J. and Rao C. N. R. (1984) A ionization spectroscopy in surface analysis and laser desorption Convenient route for the synthesis of complex metal oxides mass spectrometry. Microchim. Acta 124, 273–281. employing solid solution precursors. Inorg. Chem. 23, 1206– Zuilen M. A., Lepland A. and Arrhenius G. (2002) Reassessing the 1210. evidence for the earliest traces of life. Nature 418, 627–630. Vidyasagar K., Gopalakrishnan J. and Rao C. N. R. (1985) Synthesis of complex metal-oxides using hydroxide, cyanide, Associate editor: Christian Koeberl